WO2017168967A1 - Device for determining data analysis method candidate - Google Patents

Abstract

The purpose of the present invention is to automatically recommend an analysis algorithm for analysis target data regardless of whether a source code or an intermediate code is present. A device for determining a data analysis method candidate according to the present invention determines an analysis method candidate for analysis target data to be subjected to data analysis, and includes: an analysis example storing unit 3 that stores, as an analysis example, data in which a data attribute and an analysis method are associated with each other, for each of a plurality of previously analyzed data items; an analysis target data storing unit 2 that stores data attribute information about the analysis target data; and an analysis method candidate determining unit 4 that calculates a data attribute similarity degree as the degree of similarity between the data attribute of the analysis target data and the data attribute of each of the previously analyzed data items, and determines, as an analysis method candidate for the analysis target data, at least one analysis method among analysis methods used for the previously analyzed data items on the basis of data attribute similarity degree.

Description

Data analysis method candidate decision device

This invention relates to a technique for determining data analysis method candidates.

In order to analyze data, it is necessary to select an appropriate data analysis method according to the characteristics and meaning of the data. At present, engineers specializing in data analysis methods called data scientists recommend data analysis methods. In recent years, due to an increase in the number of devices connected to the Internet, data collected via the Internet has increased explosively. Therefore, there is an increasing need for data analysis engineers who analyze these data. However, the development of data analysis engineers has not progressed, and there are many data that have been collected but not effectively used.

In order to solve the problem of lack of data analysis engineers, a mechanism to recommend data analysis methods mechanically is necessary. As a technology in the related field, Patent Document 1 discloses a software analysis device that selects software parts to be reused or changed at the same time when developing a derivative product based on past development results and change results of software products. Yes. In the software analysis apparatus of Patent Document 1, when a user selects a software component that is converted into a source code, a software component that is considered to be used simultaneously with the software component is extracted based on the distance between the software components. To present.

Patent Document 2 discloses an information processing device that recommends a source code. The information processing apparatus of Patent Document 2 converts the source code of a program under development into intermediate code, extracts similar intermediate code from the intermediate code stored in the database, and extracts similar intermediate code source code. Recommendation to.

JP 2010-113449 A JP 2013-3664 A

However, the technique of Patent Document 1 has a problem that it cannot be used unless there is a software component that has been source-coded. In addition, since the software parts to be reused are selected using only the distance between the parts of the software parts, the reusable software parts cannot be selected based on the similarity of the analysis target data. there were.

In Patent Document 2, although there is no limitation on the language type of the source code, there is a problem that the source code cannot be recommended if there is no intermediate code generated from the source-coded program.

The present invention has been made in view of the above-described problems, and an object of the present invention is to determine analysis method candidates for data to be analyzed regardless of the presence or absence of source code or intermediate code.

A data analysis method candidate determination device according to the present invention is a data analysis method candidate determination device that determines an analysis method candidate of analysis target data to be subjected to data analysis, and a plurality of analyzed data that have been subjected to data analysis in the past An analysis case storage unit that stores data associated with data attributes and analysis methods as analysis cases, an analysis target data storage unit that stores data attribute information for the analysis target data, and the analysis target data A data attribute similarity that is a similarity between the data attribute of the analyzed data and the data attribute of the analyzed data is calculated, and at least one of the analysis methods of the analyzed data is analyzed based on the data attribute similarity An analysis method candidate determination unit that determines the analysis method candidate of the target data.

A data analysis method candidate determination device according to the present invention is a data analysis method candidate determination device that determines an analysis method candidate of analysis target data to be subjected to data analysis, and a plurality of analyzed data that have been subjected to data analysis in the past An analysis case storage unit that stores data associated with data attributes and analysis methods as analysis cases, an analysis target data storage unit that stores data attribute information for the analysis target data, and the analysis target data A data attribute similarity that is a similarity between the data attribute of the analyzed data and the data attribute of the analyzed data is calculated, and at least one of the analysis methods of the analyzed data is analyzed based on the data attribute similarity An analysis method candidate determination unit that determines the analysis method candidate of the target data. Since the analysis method candidates are determined based on the data attribute similarity, the analysis method candidates can be determined without the source code of each analysis method.

The objects, features, aspects and advantages of the present invention will become more apparent from the following detailed description and the accompanying drawings.

1 is a block diagram showing a configuration of a data analysis technique candidate determination device according to Embodiment 1. FIG. It is a figure which illustrates a data attribute. 2 is a diagram illustrating a hardware configuration of a data analysis technique candidate determination device according to Embodiment 1. FIG. 4 is a flowchart showing an operation of the data analysis technique candidate determination device according to the first embodiment. It is a flowchart which shows the process in step S15 of FIG. It is a figure which shows the example of a setting of a distance evaluation axis. 6 is a block diagram illustrating a configuration of a data analysis technique candidate determination device according to Embodiment 2. FIG. 10 is a flowchart showing the operation of the data analysis technique candidate determination device according to the second embodiment. It is a flowchart which shows operation | movement of an evaluation acquisition part. FIG. 10 is a block diagram illustrating a configuration of a data analysis technique candidate determination device according to a modification of the second embodiment. 10 is a flowchart showing an operation of a data analysis technique candidate determination device according to a modification of the second embodiment. FIG. 10 is a block diagram illustrating a configuration of a data analysis technique candidate determination device according to a third embodiment. 10 is a flowchart showing the operation of the data analysis technique candidate determination device according to the third embodiment. 10 is a flowchart showing the operation of the data analysis technique candidate determination device according to the third embodiment. It is a figure which shows the function flowchart A. It is a figure which shows the function flowchart B. FIG. 10 is a block diagram illustrating a configuration of a data analysis technique candidate determination device according to a fourth embodiment. 10 is a flowchart showing the operation of the data analysis technique candidate determination device according to the fourth embodiment. It is a flowchart which shows operation | movement of the existing data utilization proposal part in FIG.18 S19. FIG. 10 is a block diagram illustrating a configuration of a data analysis technique candidate determination device according to a fifth embodiment. 10 is a flowchart showing the operation of the data analysis technique candidate determination device according to the fifth embodiment. It is a flowchart which shows operation | movement of the analysis technique review proposal part in FIG.20 S20.

<A. Embodiment 1>
<A-1. Configuration>
FIG. 1 is a block diagram illustrating a configuration of a data analysis technique candidate determination device 11 according to the first embodiment. The data analysis method candidate determination device 11 is a device that determines an analysis method candidate of analysis target data to be analyzed and recommends it to the user. The data analysis method candidate determination device 11 includes an analysis target data storage unit 2, an analysis case storage unit 3, and an analysis method candidate determination unit 4. However, the components of the data analysis method candidate determination device 11 are not only provided in one device, but are also distributed and arranged in a plurality of devices, and the plurality of devices are connected to each other by a network such as the Internet, The data analysis method candidate determination device 11 as a system as a whole may be configured.

The data analysis technique candidate determination device 11 can use the input unit 5 and the output unit 6. The input unit 5 is an input interface for inputting a command or search condition from the user to the data analysis technique candidate determination device 11. The output unit 6 is an output interface that outputs the determination result of the analysis method candidate by the analysis method candidate determination unit 4 to the user. In FIG. 1, the input unit 5 and the output unit 6 are shown as different configurations from the data analysis method candidate determination device 11, but the data analysis method candidate determination device 11 may include these.

The analysis target data storage unit 2 is configured by a recording medium such as an HDD (Hard Disk Drive) or SD, and stores analysis target data to be analyzed and data attributes of the analysis target data. The data to be analyzed by the data analysis method candidate determination device 11 includes temperature, humidity, vibration, speed, acceleration, pressure, amount of solar radiation, distance, weight, current, voltage, electric energy, number of revolutions, Time series data such as numbers, device usage history, access logs, mobile GPS data, discrete data such as weather observations or weather forecasts, reports, inspection records, work history, forms, plans, etc. Includes document data or demographic or white paper statistical data. The data to be analyzed is data that should be analyzed from now on, but in addition to this, analyzed data that has been analyzed in the past and data analysis results newly created by data analysis, prediction, estimation, etc. are analyzed. It may be stored in the target data storage unit 2. The analysis target data storage unit 2 may include data that has not been analyzed in the past but can be used and data attributes of the data. The analysis target data storage unit 2 only needs to store the data attribute of the analysis target data, and the analysis target data itself does not necessarily have to be stored. Examples of analysis target data in which the analysis target data itself is not stored in the analysis target data storage unit 2 include open data provided by local governments, data submitted to an SNS (Social Network System), or data analysis method candidate determination device 11 There is data that is distributed and stored in a cloud environment that can be accessed from.

FIG. 2 is a diagram illustrating data attributes. FIG. 2 shows data attributes for data A, data B, and data C, respectively. Data attributes represent data characteristics, and include, for example, data acquisition intervals, data acquisition methods, distinction between actual values, predicted values, and processed values, data types, related data, and related devices. In addition, access authority for data may be used as a data attribute.

The analysis case storage unit 3 includes a recording medium such as an HDD (Hard Disk Drive) or SD. The analysis case storage unit 3 stores, as analysis cases, data in which data attributes and analysis techniques are associated with analyzed data that has been subjected to data analysis in the past. The analysis example does not need to be an analysis example created by the data analysis method candidate determination device 11, but is an existing analysis example, a publicly known example based on literature, a trial application example at the examination stage, a non-adoption example, or an analysis method change example Etc. are desirable. Further, the analysis example may include user evaluation information for the analysis method. In each analysis case, the analysis method may be described in source code, or may be described in intermediate code that can be executed by the program. Alternatively, it may be described by a name such as “regression analysis” or “k-means method”. Or you may describe by the hierarchical structure which consists of a high-order concept, a middle concept, and a low-order concept like "statistical analysis-> clustering-> k-means method". Alternatively, it may be described as an ID.

The analysis method candidate determination unit 4 selects an analysis method to be used for data analysis of the analysis target data from the analysis methods used in the past analysis examples, and determines the analysis method candidate. The analysis method candidate determined here is output from the output unit 6 in, for example, a text format and recommended to the user. Alternatively, a combination of representative past cases with analysis method candidates may be output in a list format and recommended to the user. In this case, the user can easily understand examples or features of the analysis technique candidates.

FIG. 3 is a diagram illustrating a hardware configuration of the data analysis technique candidate determination apparatus 11. The data analysis technique candidate determination device 11 includes a processor 20, a memory 21, and a recording medium 22. The analysis technique candidate determination unit 4 is realized as a function of the processor 20 by executing a software program stored in a memory 21 such as a RAM (Random Access Memory) by a processor 20 such as a CPU (Central Processing Unit). To do. However, these may be realized in cooperation with a plurality of processors. Note that the analysis technique candidate determination unit 4 may be realized by a signal processing circuit that realizes the operation by a hardware electric circuit. As a concept combining the software analysis method candidate determination unit 4 and the hardware analysis method candidate determination unit 4, the word “processing circuit” can be used instead of the word “part”.

<A-2. Operation>
FIG. 4 is a flowchart showing the operation of the data analysis technique candidate determination apparatus 11. First, the user selects analysis target data and analysis purpose via the input unit 5 (step S11). For the analysis target data, for example, a list of data stored in the analysis target data storage unit 2 may be displayed, and the user may select from the list, or the user inputs new analysis target data using an electronic file or the like. You may be able to do it. When analysis target data is newly input, the data is stored in the analysis target data storage unit 2.

For the purpose of analysis, for example, a list such as a pull-down menu may be displayed, and the user may select from the list, or the user may be able to input with a character string. The analysis purpose selected by the user here is stored in the analysis object data storage unit 2. Further, the analysis purpose is not limited to one, and there may be a plurality of analysis purposes. Here, “television viewing data” and “viewer viewing preference analysis” will be described as analysis target data and analysis purpose examples, respectively.

Next, the analysis target data is read from the analysis target data storage unit 2 to the analysis technique candidate determination unit 4 (step S12). That is, television viewing data collected from each television terminal is read as analysis target data.

Subsequently, the data attribute and analysis purpose of the analysis target data are read from the analysis target data storage unit 2 to the analysis technique candidate determination unit 4 (step S13). That is, as data attributes of “TV viewing data” that is analysis target data, for example, the data acquisition interval, the location of the data acquisition device, and the owner information of the data acquisition device are read. Read "Analysis".

Subsequently, an analysis case having a data attribute that is the same as or similar to the analysis target data or an analysis purpose that is the same or similar to the analysis target data is read from the analysis case storage unit 3 to the analysis technique candidate determination unit 4 (step S14). For example, analysis examples similar to the data to be analyzed “TV viewing data” include “TV region audience rating survey”, “regional favorite talent analysis”, “popular movie genre survey”, “ “Power usage survey” or “Production efficiency analysis in factories”. Examples of similar analysis purposes include "Internet browsing history analysis", "Product purchase status analysis", "Visit store analysis", "Point card holding status analysis", "Public transport ride history", Or “visit facility analysis during travel”.

Subsequently, the analysis method candidate determination unit 4 determines analysis method candidates for the analysis target data (step S15). Detailed processing contents in step S15 will be described later.

Finally, the analysis technique candidate created in step S15 is output to the output unit 6 and recommended to the user (step S16), and the process ends.

FIG. 5 is a flowchart showing analysis method candidate determination processing by the analysis method candidate determination unit 4 in step S15 of FIG. First, the data attribute similarity between the analysis target data and the analyzed data is calculated for the analysis example read in step S14 of FIG. 4 (step S151). The processing will be specifically described using “public transit history” data as an example of analyzed data of an analysis example. Data attribute of “TV viewing data” that is analysis target data designated by the user, and “Traffic IC card boarding history” data or “GPS data” used for analyzing analyzed data “Boarding history of public transportation” The data attribute similarity Sz is calculated with respect to data attributes such as "the public transportation route estimated from". The data attribute similarity Sz is calculated by the following equation, for example.

However, N is the number of items registered as data attributes, Lmaxi is the maximum distance of the i-th data attribute item, and Li is the distance of the i-th data attribute item. For example, a distance evaluation axis is set for each data attribute item, and the distance Li of the i-th data attribute item is calculated using the distance evaluation axis.

Fig. 6 shows an example of setting the distance evaluation axis. For example, regarding the data acquisition interval, the distance is set to 10 if at least one of the data to be analyzed and the analyzed data is irregular. If the data acquisition interval of the analyzed data is shorter than the data acquisition interval of the analysis target data, the distance is set to zero. The distance is set to 5 if one acquisition interval of analysis target data and analyzed data is 100 times or more the other acquisition interval. Regarding the data acquisition method, for example, the distance is 0 for the same method, the distance is 2 if one is a log and the other is a terminal input, and the distance is 1 if both are sensor logs but the sensor types are different. For example, if both are actual values, the distance is 0, if one is the actual value and the other is the predicted value, the distance is 20, and if both are the predicted values, the distance is 100. Thus, the distance evaluation axis may be set on a rule basis for each data attribute item, or may be set by a mathematical expression. Further, the number of rules need not be limited, and the maximum value of the distance may be provided for each evaluation axis. Among the distance evaluation axes set as shown in FIG. 6, the maximum distance is determined as the maximum distance. In FIG. 6, the case where the distance has only a positive value is described. However, there may be a distance that takes a negative value, or a two-dimensional value or more may be taken without taking a one-dimensional value.

Subsequently, for the analysis example for which the data attribute similarity is calculated in step S151, an analysis purpose similarity Sp with the analysis target data is calculated (step S152). For example, the analysis objective of the analysis target data and the analysis objective of the analyzed data are compared with character strings, and the similarity is calculated as the analysis objective similarity Sp. The analysis target similarity Sp can be obtained by using, for example, a cosine similarity or a Levenshtein distance. For example, when the analysis target similarity Sp between the analysis target character string A of the analysis target data and the analysis target character string B of the analyzed data is obtained as the cosine similarity, it is calculated by the following equation.

However, A · B is the inner product of character string A and character string B, | A | is the distance of character string A, and | B | is the distance of character string B.

The analysis target character string A is “analysis of viewer's viewing preference” and the analysis target character string B of the analyzed data is “popular movie genre survey”. A calculation method will be described. When keywords are extracted by decomposing the character string A into word levels, “viewing, viewers, preferences, analysis” is obtained, and “popularity, movie, genre, survey” is obtained from the character string B as well. At this time, similar words may be linked and the keyword of the character string B may be “preference, movie, genre, analysis” like “preference = popular” or “analysis = survey”. A similar word database in which similar words are defined can be provided in the analysis target data storage unit 2 or the analysis example storage unit 3, and similar words can be linked by referring to the similar word database.

When the character strings A and B are displayed as vectors, A: (viewing, audience, preference, analysis, movie, genre) = (2,1,1,1,0,0), B: (viewing, audience, preference, analysis) , Movie, genre) = (0, 0, 1, 1, 1, 1).

Also, the analysis target similarity Sp is calculated as follows.

As another example, when the analysis purpose is described in source code or intermediate code, the processing procedure shown in the source code or intermediate code can be changed using a technique such as UML (Unified Modeling Language) or a function flowchart. The analysis purpose similarity Sp may be calculated from the similarity of the processing procedures. Hereinafter, a method for calculating the analysis object similarity Sp will be described with reference to the function flowchart A shown in FIG. 15 and the function flowchart B shown in FIG.

The function flowchart A indicates that steps S21 to S26 are executed in order. Step S21 is a step of inputting X, step S22 is a step of substituting X / 5 for Y, step S23 is a step of outputting Y, step S24 is a step of inputting Z, and step S25 is a step of substituting Y × Z into A Step S26 is a step for outputting Y.

Function flowchart B shows that steps S31 to S33 are executed in order. Step S31 is a step of inputting X, step S32 is a step of a subroutine relating to X, and step S33 is a step of outputting Y. Step S32 of the subroutine relating to X is step S34 in which X / 5 is substituted for Y.

Suppose that in each of these two function flowcharts A and B, the matching rate of processing procedures is defined by the number of matching processing procedures relative to the total number of processing procedures. When only the input / output process and the arithmetic process are to be counted in the processing procedure, the coincidence rate is calculated as follows.

When the magnitude of the number of consecutive matching processing procedures is added to this matching rate, the analysis target similarity degree Sp can be expressed by the following equation, for example.

In addition, when the analysis purpose is described in a hierarchical structure composed of a superordinate concept, a middle concept, and a subordinate concept, the analysis target similarity of each of the superordinate concept, the middle concept, and the subordinate concept is expressed by Equation (6). You may calculate and take the average. Alternatively, an ID number considering the similarity of the method is assigned to each of the upper concept, middle concept, and lower concept options in advance, and the analysis target similarity Sp is based on the difference between the numbers obtained by combining the ID numbers. You may ask for.

For example, if the maximum value of the ID number is “9-9-99”, the analysis purpose in which the ID number of the superordinate concept—medium concept—subordinate concept is represented by “1-0-01” and the superordinate concept—medium The analysis object similarity Sp with the analysis object whose ID number is “1-0-02” can be calculated as follows.

For the purpose of analysis where the ID number of the superordinate concept-intermediate concept-subordinate concept is represented by “1-0-01”, the ID number of the superordinate concept-intermediate concept-subordinate concept is “5- The analysis object similarity Sp with the analysis object represented by “0-01” can be calculated as follows.

The formula for calculating the analysis target similarity degree Sp described above is merely an example. Therefore, it is possible to make modifications such as weighting a specific condition or performing a correction operation such as tilting when there is a bias in the average value of calculation results due to a difference in the calculation method of the analysis target similarity. .

In addition, if there are cases with different description methods for analysis purposes, cases that represent multiple cases are extracted, and only the representative cases are assigned analysis purposes for all analysis purpose description methods, so that In particular, comparison for analysis purposes may be made possible.

Subsequently, based on the data attribute similarity Sz and the analysis target similarity Sp, an overall similarity S between the analysis target data and the analyzed data is calculated (step S153). The total similarity S is calculated by the following formula, for example.

Subsequently, it is confirmed whether there is any other analyzed data for which the total similarity is not calculated (step S154). If there is analyzed data for which the total similarity is not calculated, the process returns to step S151, and the processes from step S151 to step S153 are executed on the analyzed data. When the calculation of the similarity is completed for all analyzed data, the process proceeds to step S155.

In step S155, an average similarity is calculated for each analysis method from the total similarity of all analysis cases read in step S14 of FIG. For example, in the analysis example read in step S14 in FIG. 4, analysis such as “regression analysis”, “k-means method”, “behavior model base reasoning”, “behavior model base reasoning and queuing simulation”, “neural network”, etc. Suppose that the technique was used. At this time, the average similarity Sav with respect to the “regression analysis” is calculated by the following equation, for example.

However, N _times indicates the number of cases including “regression analysis” as the data analysis method, and ΣS _times indicates the sum of the total similarities of cases including “regression analysis” as the data analysis method. Although the arithmetic average is used in the above example, the average similarity may be calculated using various other averages such as a geometric average, a harmonic average, and a weighted average.

When a plurality of analysis methods are used in one case, the average similarity may be calculated while maintaining a combination of a plurality of analysis methods. Alternatively, after calculating the average similarity as a single method, the average similarity may be calculated again for the data analysis methods used in combination only for the method having a high average similarity.

Finally, analysis method candidates for the analysis target data are determined (step S156). Here, the analysis method with the highest average similarity may be set as the analysis method candidate, or a plurality of analysis methods may be set as the analysis method candidates in descending order of the average similarity. When outputting the analysis method candidates in step S16 of FIG. 4, in addition to the analysis method candidates, the average similarity, the number of analysis cases including the analysis method candidates, or the appearance frequency of the analysis purpose using the analysis method candidates Etc. may be output together.

<A-3. Effect>
The data analysis technique candidate determination device 11 according to the first embodiment stores an analysis example in which data associated with a data attribute and an analysis technique is stored as an analysis example for each of a plurality of analyzed data subjected to data analysis in the past. The storage unit 3, the analysis target data storage unit 2 that stores data attribute information, and the data attribute similarity that is the similarity between the data attribute of the analysis target data and the data attribute of the analyzed data is calculated. And an analysis method candidate determination unit 4 that determines at least one analysis method as an analysis method candidate for the analysis target data from among the analysis methods of the analyzed data based on the data attribute similarity. Therefore, even if there is no source code for each analysis method, analysis method candidates can be determined with reference to analysis cases having similar data attributes.

The analysis case storage unit 3 stores analysis purpose information for each of a plurality of analyzed data, and the analysis target data storage unit 2 stores analysis purpose information of the analysis target data to determine analysis method candidates. The unit 4 calculates the similarity between the analysis purpose of the analysis target data and the analysis purpose of the analyzed data as the analysis target similarity, and combines the analysis target data and the analyzed data based on the analysis target similarity and the data attribute similarity. The similarity is calculated, and at least one analysis method is determined as an analysis method candidate for the analysis target data from the analysis methods of the analyzed data based on the total similarity. Therefore, even if there is no source code for each analysis method, analysis method candidates can be determined with reference to analysis cases having similar data attributes and analysis purposes.

Further, the data attributes of the analyzed data and the analysis target data include at least one of a data acquisition interval, a data acquisition method, and an actual value, a predicted value, or a processed value. By determining the analysis method candidates based on the similarity of these data attributes, the analysis method candidates can be determined without the source code of each analysis method.

Further, the analysis method candidate determination unit 4 calculates the analysis target similarity based on the analysis target character string of the analysis target data and the analysis target character string of the analyzed data. By comparing the character strings and calculating the analysis target similarity, and determining the analysis method candidate based on the analysis target similarity, the analysis method candidate can be determined without the source code of each analysis method. it can.

Also, the analysis technique candidate determination unit 4 calculates the analysis object similarity based on the analysis purpose of the analysis target data described in the hierarchical structure and the analysis purpose of the analyzed data described in the hierarchical structure. By comparing the similarities between the analytical objectives set in advance for each hierarchy, calculating the analytical objective similarity, and determining the analytical technique candidates based on the analytical objective similarity, there is no source code for each analytical technique. Also, analysis method candidates can be determined.

When the analysis purpose of the analysis target data and the analysis purpose of the analyzed data are described in the source code or the intermediate code, the analysis method candidate determination unit 4 is indicated in the source code or the intermediate code for the analysis target data. The similarity between the processing procedure and the processing procedure indicated in the analysis-purpose source code or intermediate code of the analyzed data is calculated as the analysis-object similarity based on the matching rate or the continuity of the matching processing procedures. Analytical purpose is described in source code or intermediate code by calculating analytical target similarity based on the matching rate of processing procedures or continuity of matching processing procedures and determining analysis method candidates based on the analytical target similarity Even in such a case, analysis method candidates can be determined.

The analysis method candidate determination unit 4 calculates the average value of the overall similarity between the analyzed data using the analysis method and the analysis target data for each analysis method, and selects the average based on the average value of the overall similarity The analysis method is determined as a candidate analysis method. Therefore, analysis method candidates can be determined without the source code of each analysis method.

<B. Second Embodiment>
<B-1. Configuration>
FIG. 7 is a block diagram illustrating a configuration of the data analysis technique candidate determination device 12 according to the second embodiment. In addition to the configuration of the data analysis technique candidate determination device 11 according to Embodiment 1, the data analysis method candidate determination apparatus 12 newly includes an evaluation acquisition unit 7 and a recommended case storage unit 8.

The recommended case storage unit 8 is composed of a recording medium such as an HDD (Hard Disk Drive) or SD, and stores recommended case data. The recommended case data is data in which analysis method candidates determined by the analysis method candidate determination unit 4 in the past are associated with the analysis target data and the analysis purpose.

The evaluation acquisition unit 7 acquires evaluation information for the analysis technique candidate input by the user via the input unit 5, and adds the evaluation information to the corresponding recommended case stored in the recommended case storage unit 8. That is, the recommended case storage unit 8 stores recommended cases made up of analysis target data, analysis purposes, and analysis method candidates, and evaluation information for the recommended cases in association with each other. The evaluation acquisition unit 7 is realized as a function of the processor 20 by the processor 20 illustrated in FIG. 3 executing a software program stored in the memory 21.

<B-2. Operation>
FIG. 8 is a flowchart showing the operation of the data analysis technique candidate determination device 12. Steps S11 to S16 are the same as those in the first embodiment and have already been described with reference to FIG. The analysis method candidate determination unit 4 determines the analysis method candidate (step S15), and outputs the analysis method candidate to the output unit 6 (step S16), and associates the analysis target data, the analysis purpose, and the analysis method candidate. Data (recommended case) is stored in the recommended case storage unit 8 (step S17).

FIG. 9 is a flowchart showing the operation of the evaluation acquisition unit 7. This flow is performed only when a recommended case is stored in the recommended case storage unit 8. First, the evaluation acquisition unit 7 determines a recommended case to which evaluation information is to be added (step S71). For example, a screen displaying a list of all recommended cases stored in the recommended case storage unit 8 may be displayed, and the user may select a recommended case from the screen. In addition, the user may be allowed to input conditions such as analysis target data or analysis purpose, and the recommended cases may be specified or narrowed down based on the input conditions. Moreover, the recommended case to which evaluation information is not yet added may be extracted from the recommended case storage unit 8 and presented to the user, and the user may select it.

Next, among the plurality of analysis method candidates recommended in the recommended case determined in step S71, the analysis method candidate actually used by the user is specified (step S72). When the user uses a plurality of analysis technique candidates, the plurality of analysis technique candidates are specified. Here, for example, a list screen of a plurality of analysis method candidates is displayed, and analysis method candidates actually used by the user are selected from the list screen.

Next, user evaluation information is acquired for the analysis method candidate specified in step S72 (step S73). The user evaluation information is obtained by causing the user to input from the input unit 5. The evaluation information includes supplementary information such as analysis accuracy, user's personal feeling, execution time, and the like. Alternatively, the analysis method candidate that provides the most desirable result may be selected by the user from a list screen of a plurality of analysis method candidates. Alternatively, instead of selecting the most desirable one, the ranks may be input to the analysis method candidates in the order in which desirable results are obtained.

Moreover, in addition to the information related to good evaluation as described above, information related to bad evaluation may be acquired. For example, if there is an analysis method candidate that has been used by the user but has not been adopted as a result due to some problem, for example, a problem related to the analysis method candidate may be input. In addition, it is possible to input analysis methods candidates that are not actually used by the user. Further, supplementary information such as assignments may be selected from answers prepared in advance or may be freely input.

The evaluation acquisition unit 7 assigns the evaluation information thus acquired to the recommended case and stores it in the recommended case storage unit 8 (step S74).

Furthermore, the evaluation acquisition unit 7 adds, to the analysis case storage unit 3, a recommended case related to an analysis method candidate to which desirable evaluation information is assigned among recommended cases to which evaluation information is assigned (step S <b> 75). ). For example, among the analysis method candidates “regression analysis” and “k-means method” for the analysis target data “TV viewing data” and the analysis purpose “analysis of viewer viewing preference”, desirable evaluation for “regression analysis” When information is acquired and undesirable evaluation information is acquired for the “k-means method”, the analysis target data “TV viewing data”, the analysis purpose “analysis of viewer viewing preferences”, the analysis method “regression analysis” Is added to the analysis case storage unit 3 as a new analysis case. When desirable evaluation information is obtained for a plurality of analysis methods, all the analysis methods that have obtained desirable evaluation information are added to the analysis case storage unit 3 as described above. In this way, an analysis example obtained with desirable evaluation information is added, and the analysis method candidate is determined using the analysis example, thereby improving the determination accuracy of the analysis method candidate.

<B-3. Modification>
FIG. 10 is a block diagram illustrating a configuration of a data analysis technique candidate determination device 13 according to a modification of the second embodiment. The data analysis technique candidate determination device 13 includes an attribute adding unit 9 in addition to the configuration of the data analysis technique candidate determination device 12. The configuration of the data analysis method candidate determination device 13 other than the attribute adding unit 9 is the same as that of the data analysis method candidate determination device 12.

The attribute adding unit 9 analyzes the reason for non-adoption of the analysis technique candidate acquired by the evaluation acquiring unit 7, and sets the data attribute corresponding to the reason for non-adoption to all the data attributes stored in the analysis target data storage unit 2. It adds as a new data attribute item of analysis object data. At this time, the attribute adding unit 9 may notify the user such as a system administrator of the added data attribute item through the output unit 6 and prompt the user to input the data attribute related to the added data attribute item. Also, the distance evaluation axis for calculating the data attribute similarity for the added data attribute item may be urged to be input to the user in the same manner as the data attribute. The user can input these data attributes or distance evaluation axes to the data analysis technique candidate determination device 13 through the input unit 5. The attribute adding unit 9 is realized as a function of the processor 20 when the processor 20 illustrated in FIG. 6 executes a software program stored in the memory 21.

FIG. 11 is a flowchart showing the operation of the attribute adding unit 9 in the data analysis technique candidate determination device 13. This flow is executed when the reason for not adopting the analysis method candidate is stored in the recommended case storage unit 8.

First, a recommended case to which evaluation information is given is extracted from the recommended case storage unit 8 (step S81).

Next, the reason for non-adoption is extracted for the analysis method candidate that is not used in the recommended case extracted in step S81 (step S82).

Subsequently, the reason for non-employment extracted in step S82 is analyzed (step S83). As an analysis method, frequency analysis by keyword extraction, simple statistics, or the like can be used.

Finally, the data attribute item corresponding to the analyzed non-recruitment reason is added as the data attribute item of the analysis target data stored in the analysis target data storage unit 2 (step S84). For example, as a result of analyzing the reason for non-adoption in step S83, if it is found that keywords such as “execution time is long” and “processing is heavy” are many as reasons for non-adoption, “calculation amount”, “execution time per unit amount” To the data attribute.

As described above, according to the data analysis method candidate determination device 13, by adding the data attribute corresponding to the reason for not using the analysis method candidate, the analysis method candidate determination unit 4 determines the analysis method candidate more finely. It becomes possible to judge attribute similarity. Therefore, it is possible to improve the determination accuracy of the analysis method candidate.

<B-4. Effect>
In addition to the configuration of the data analysis method candidate determination device 11 according to the first embodiment, the data analysis method candidate determination device 12 according to the second embodiment includes an evaluation acquisition unit 7 that acquires user evaluation information for the analysis method candidate; And a recommended case storage unit 8 that stores data associating data attributes of the analysis target data, analysis method candidates of the analysis target data, and evaluation information for the analysis method candidates as recommended cases. As described above, if the determination result of the analysis method candidate is stored as a recommended case, the determination accuracy of the analysis method candidate can be improved by using, for example, the recommended case that has obtained desirable evaluation information as the analysis case.

In addition to the configuration of the data analysis technique candidate determination device 12 according to the second embodiment, the data analysis method candidate determination apparatus 13 according to the modification of the second embodiment analyzes from the evaluation information acquired by the evaluation acquisition unit 7. An attribute adding unit 9 is provided for extracting the reason for non-adoption of method candidates and adding an item corresponding to the reason for non-adoption to the data attribute item. Therefore, when the analysis method candidate determination unit 4 determines the analysis method candidate, the data attribute similarity can be determined more finely, so that the determination accuracy of the analysis method candidate can be improved.

<C. Embodiment 3>
<C-1. Configuration>
FIG. 12 is a block diagram illustrating a configuration of the data analysis technique candidate determination device 14 according to the third embodiment. The data analysis method candidate determination device 14 includes a model change proposing unit 10 in addition to the configuration of the data analysis method candidate determination device 11 according to the first embodiment.

The model change proposing unit 10 proposes a physical model change such as correction or addition of a physical model when the analysis method candidate determined by the analysis method candidate determining unit 4 includes a physical model-based analysis method. Here, the physical model-based analysis method indicates all data analysis methods using a physical model based on data or design information, such as a device model, a failure model, a behavior model, a correlation model, or a user model. The physical model may be described in a document format such as a parameter sheet, may be described in a chart format such as an FTA (Fault Tree Analysis) diagram, a fault tree, or an electric circuit diagram, an equation of motion or a bathtub curve, etc. Or a machine language such as assembler or source code. The model change proposing unit 10 is realized as a function of the processor 20 by the processor 20 illustrated in FIG. 3 executing a software program stored in the memory 21.

The analysis case storage unit 3 stores analysis target data, an analysis purpose and data attribute of the analysis target data, and an analysis technique as an analysis case. Furthermore, when the analysis method is a physical model-based analysis method, change information of the physical model is also stored as an analysis example. Specifically, when a user makes a change (addition, modification) to a physical model and then performs data analysis using the changed physical model, the changed physical model actually used for data analysis Not only the model but also the physical model before the change is stored in the analysis case storage unit 3 as change information.

The configuration of the data analysis technique candidate determination device 14 other than that described above is the same as the configuration of the data analysis method candidate determination device 11 according to the first embodiment.

<C-2. Operation>
FIG. 13 is a flowchart showing the operation of the data analysis technique candidate determination device 14. Steps S11 to S15 and S16 are the same as in the first embodiment, but differ from the first embodiment in that a new step S18 is added between steps S15 and S16. When the analysis method candidate of the analysis target data is determined by the analysis method candidate determination unit 4 (step S15), when the analysis method candidate includes a physical model-based analysis method, the model change proposing unit 10 proposes a change of the physical model. (Step S18).

FIG. 14 is a flowchart showing the operation of the model change proposing unit 10 in step S18 of FIG. This flow is executed only when physical model change information is stored in the analysis case storage unit 3.

First, it is determined whether the analysis method candidate determined by the analysis method candidate determination unit 4 in step S15 in FIG. 13 includes a physical model base analysis method (step S181). If the physical model base analysis method is not included, the process of the model change proposing unit 10 is terminated. If the physical model-based analysis method is included, the process proceeds to step S182.

In step S182, the analysis example stored in the analysis example storage unit 3 uses the same analysis method as the physical model database analysis method included in the analysis method candidates, and the analysis example describes the change information of the physical model To extract.

Next, it is determined whether or not the changed physical model data indicated by the change information is stored in the analysis case storage unit 3 (step S183). If the changed physical model data exists in the analysis example storage unit 3, the user is suggested to utilize the changed physical model (step S184). For example, it is assumed that an analysis method that uses the passenger model A as a physical model has been recommended as an analysis method candidate when the user previously analyzed the analysis target data “boarding history of public transportation”. On the other hand, when the user performs a data analysis using the passenger model B in which some modifications are made to the passenger model A or a new passenger model is added, the analysis example storage unit 3 includes: In addition to the analysis target data, the analysis purpose, and the analysis method actually used (passenger model B), the passenger model A before the change is recorded. Thereafter, in another data analysis, when the analysis method candidate determination unit 4 determines an analysis method that uses the passenger model A as a physical model as the analysis method candidate, the passenger model B is used instead of the passenger model A. Suggest to users.

In step S183, if the changed physical model data does not exist in the analysis example storage unit 3, a method for changing (correcting or adding) the physical model is proposed to the user. For example, if the analysis method that uses the customer model as a physical model is an analysis method candidate for the analysis purpose of “product purchase situation analysis”, the customer model is corrected to a category suitable for the product genre you want to analyze. And a method for adding a purchaser model that “a parent buys on behalf of a child” is proposed.

<C-3. Effect>
In the data analysis technique candidate determination device 14 according to the third embodiment, the analysis case data stored in the analysis case storage unit 3 is an analysis in which data analysis is performed using a physical model obtained by changing a certain physical model. For the case, information on the physical model before the change is included. The data analysis technique candidate determination device 14 includes a model change proposing unit 10 in addition to the configuration of the data analysis technique candidate determination device 11 according to the first embodiment. The model change proposing unit 10 proposes a change of the physical model when the analysis method candidate is an analysis method using a physical model and the physical model used in the analysis method candidate is the same as the physical model before the change in the analysis example. To do. Therefore, it is possible to improve the analysis accuracy regarding the physical model-based analysis method.

<D. Embodiment 4>
<D-1. Configuration>
FIG. 17 is a block diagram illustrating a configuration of the data analysis technique candidate determination device 15 according to the fourth embodiment. The data analysis technique candidate determination device 15 includes an existing data utilization proposal unit 101 in addition to the configuration of the data analysis technique candidate determination device 11 according to the first embodiment.

When the analysis target data (first analysis target data) selected by the user does not have the data attributes necessary for the execution of the analysis method determined by the analysis method candidate determination unit 4, the existing data utilization proposal unit 101 Extract the analysis target data (second analysis target data) having the necessary data attributes from the past analysis target data stored in the analysis target data storage unit 2, and propose to the user to use the second analysis target data To do. The existing data utilization proposing unit 101 is realized as a function of the processor 20 by the processor 20 illustrated in FIG. 3 executing a software program stored in the memory 21.

The analysis case storage unit 3 stores the analysis target data initially selected by the user, the analysis purpose and data attributes of the analysis target data, and the analysis technique as analysis cases. The analysis case storage unit 3 also stores analysis target data additionally selected by the user as proposed by the existing data utilization proposal unit 101 as an analysis case. The analysis target data may be stored in the analysis case storage unit 3 with a flag for each selection timing.

The configuration of the data analysis method candidate determination device 15 other than that described above is the same as the configuration of the data analysis method candidate determination device 11 according to the first embodiment.

<D-2. Operation>
FIG. 18 is a flowchart showing the operation of the data analysis technique candidate determination device 15. In the flowchart of FIG. 18, steps S11 to S15 and S16 are the same as in the first embodiment, but differ from the first embodiment in that a new step S19 is added between steps S15 and S16. When the analysis method candidate of the analysis target data is determined by the analysis method candidate determination unit 4 (step S15), the data attribute of the analysis target data acquired in step S13 is used as a data attribute necessary for executing the analysis method candidate. If it is insufficient, the existing data utilization proposing unit 101 proposes to add analysis target data (step S19).

FIG. 19 is a flowchart showing the operation of the existing data utilization proposing unit 101 in step S19 of FIG.

First, the existing data utilization proposing unit 101 uses the data attributes necessary for the analysis target data (first analysis target data) selected in step S11 of FIG. 18 to execute the analysis method candidates determined in step S15. It is judged whether it has (step S191). Here, as the case where the analysis target data does not have the necessary data attribute, the following three cases are exemplified. The first is a case where the analysis target data itself is missing. The second case is a case where the data acquisition interval of the analysis target data is rough with respect to the data acquisition interval specified as the necessary data attribute, and a sufficient analysis result cannot be obtained. The third is a case where the data acquisition method specified as the necessary data attribute does not match the data acquisition method of the analysis target, and sufficient analysis results cannot be obtained. For example, the third case corresponds to the case where the analysis target data is a processed value even though the data is directly measured by a sensor or the like.

If the analysis target data (first analysis target data) has a data attribute necessary for executing the analysis method candidate, the existing data utilization proposing unit 101 ends the process. On the other hand, when the analysis target data (first analysis target data) does not have a data attribute necessary for executing the analysis method candidate, the existing data utilization proposal unit 101 proceeds to the process of step S192.

In step S192, the existing data utilization proposing unit 101 uses an analysis method that is the same as or includes an analysis method candidate from the analysis cases stored in the analysis case storage unit 3, and the analysis purpose is the same or similar. Extract analysis cases.

Next, the existing data utilization proposing unit 101 compares the data attribute of the analyzed data in the extracted analysis example with the data attribute of the analysis target data currently selected by the user, and determines the data attribute of the analyzed data. Data attributes necessary for executing the analysis technique candidate are extracted (step S193). At this time, data access authority is set as the data attribute and the user does not have access authority, or data utilization conditions are set as the data attribute and the data is diverted by contract with the data source. Data attributes such as restricted data may be excluded from extraction. In this case, only the data attribute may be presented by giving access authority or restriction information on data diversion.

Then, if the analysis target data having the data attribute extracted in step S193 exists in the analysis target data storage unit 2, the existing data utilization proposing unit 101 determines that the analysis target data having the extracted data attribute (first 2 analysis target data), that is, the second analysis target data is added to the currently selected analysis target data (first analysis target data), and the user is proposed to perform the analysis (step S194). For example, when the user analyzes the analysis target data “the power consumption of a general household existing in A prefecture B city C town D chome” by adding the analysis target data “weekday / holiday classification of the analysis target period” Assume that the “k-means method” is presented as a method candidate, and the user decides to use the analysis method candidate. At this time, another user uses the “k-means method” in the analysis case storage unit 3 to change the analysis target data “building energy consumption” into the analysis target data “weekday / holiday classification of the analysis target period”. Suppose that there was a case in which “meteorological observation data during the analysis period” and “employee entry / exit history during the analysis period” were added and analyzed. However, it is assumed that the analysis target data “employee entry / exit history of the analysis target period in the building” indicates that secondary use of the data is impossible as a data attribute. In that case, the existing data utilization proposing unit 101 may propose to the user to additionally use the analysis target data “meteorological observation data in the analysis target period” in step S194. At this time, the existing data utilization proposing unit 101 preferably uses the analysis target data “meteorological observation data during the analysis target period” and the analysis target data “employee entry / exit history of the analysis target period”. It may be shown to the user that the secondary use of data is indicated as a data attribute of the data “history of entry / exit of employees in the analysis period”.

In the above, three cases are exemplified as cases where the analysis target data does not have the data attribute necessary for applying the analysis method candidate, and in such a case, the addition of the analysis target data is proposed. explained. However, even if the analysis target data has a data attribute necessary for applying the analysis method candidate, addition of the analysis target data may be proposed in the following cases. The first is a case where analysis target data having a necessary data attribute but a condition that does not provide the best result is selected. The second is a case where analysis is possible even with the currently selected analysis target data, but more accurate analysis results can be obtained by adding new analysis target data.

<D-3. Effect>
In the data analysis method candidate determination device 15 according to the fourth embodiment, the first analysis target data has data attributes necessary for the analysis method determined by the analysis method candidate determination unit 4 with respect to the first analysis target data. If not, an existing data utilization proposing unit 101 that proposes utilization of second analysis target data having necessary data attributes is provided. Thus, by proposing the addition of another analysis target data having data attributes necessary for the execution of the analysis method candidate, it is possible to improve the analysis accuracy when the analysis method candidate is executed.

In addition, the second analysis target data has a data attribute related to whether or not the data can be diverted. When the existing data utilization proposing unit 101 proposes utilization of the second analysis target data to the user, whether or not the analyzed data can be diverted. Information to users. Accordingly, when the second analysis target data proposed to the existing data utilization proposing unit 101 is data that cannot be diverted, the user can consider obtaining alternative data that can be diverted, and add alternative data. Thus, it is possible to improve the analysis accuracy when the analysis method candidate is executed.

<E. Embodiment 5>
<E-1. Configuration>
FIG. 20 is a block diagram illustrating a configuration of the data analysis technique candidate determination device 16 according to the fifth embodiment. The data analysis method candidate determination device 16 includes an analysis method review proposal unit 102 in addition to the configuration of the data analysis method candidate determination device 11 according to the first embodiment.

The analysis method review proposal unit 102 calculates the adoption rate for each analysis method when a case with the same or similar analysis purpose is added to the analysis case stored in the analysis case storage unit 3 and sets it in advance. If an analysis method of the employment rate that satisfies the analysis method review condition is detected, the analysis method is proposed to the user. The analysis technique review proposal unit 102 is implemented as a function of the processor 20 by the processor 20 illustrated in FIG. 3 executing a software program stored in the memory 21.

In the analysis case storage unit 3, the information of the user who registered or updated the analysis case, the information of the person in charge of the analysis case, the information of the developer or the provider of the analysis method, the current use state of the analysis case Etc. are preferably stored. The current utilization status of analysis cases may include external cases, etc. in addition to the usage status of product application, trial, or cancellation.

The configuration of the data analysis method candidate determination device 16 other than that described above is the same as the configuration of the data analysis method candidate determination device 11 according to the first embodiment.

<E-2. Operation>
FIG. 21 is a flowchart showing the operation of the data analysis technique candidate determination device 16. Steps S11 to S16 are the same as in the first embodiment, but differ from the first embodiment in that a new step S20 is added after step S16. The analysis method candidate determination unit 4 determines the analysis method candidate of the analysis target data (step S15), and when the analysis method candidate is presented to the user (step S16), the analysis method and the average similarity for each analysis method are proposed to review the analysis method. The analysis method review proposal unit 102 determines whether or not the analysis method review proposal is necessary for the past analysis cases stored in the analysis case storage unit 3 (step S20).

FIG. 22 is a flowchart showing the operation of the analysis technique review proposing unit 102 in step S20 of FIG.

First, the analysis method review proposing unit 102 receives the analysis purpose and the average similarity for each analysis method calculated by the analysis method candidate determining unit 4 in step S15 of FIG. 21 (step S201). Subsequently, it is determined whether or not the analysis method has reached the review standard (step S202). The review criteria are, for example, that the average similarity exceeds or is below the threshold. In addition, the analysis method review proposal unit 102 holds the reception history of the average similarity for each analysis method for a certain period or a certain number of receptions, and the reception rate for each analysis method exceeds the threshold, or the reception When the correlation between the date and the average similarity shows an increasing tendency or a decreasing tendency for a certain period or more, it may be determined that the review standard has been reached. If the analysis technique does not reach the review criteria, the analysis technique review proposal unit 102 ends the process. On the other hand, if the analysis technique has reached the review criteria, the analysis technique review proposal unit 102 proceeds to the process of step S203.

In step S203, the analysis technique review proposing unit 102 extracts a past analysis case that is the same as or similar to the analysis purpose received in step S201 from the analysis case storage unit 3. At this time, the number of extractions may be limited such that N cases (for example, N = 1000) are extracted from the new cases of registration date and update date. In addition, the extraction period may be limited such that only the analysis cases in the N years (for example, N = 5) with the latest registration date or update date are extracted.

Next, the adoption rate of the analysis technique used in the extracted analysis case is calculated (step S204). The adoption rate P can be calculated by, for example, P = Nx / N. However, N is the number of extractions, and N _{x is the} number of methods X adopted. At this time, when the current utilization state of the analysis case is stored in the analysis case storage unit, the analysis case may be weighted according to the utilization state. That is, the weight is increased for the analysis example that has been applied to the product, and the weight is decreased for the analysis example that has been commercialized. Or you may weight according to the registration date of an analysis example, or an update date. That is, an analysis case with a new registration date or update date / time has a higher weight, and an analysis example with a new registration date / update date / time has a lower weight.

Next, if there is an analysis method whose adoption rate meets the analysis method review condition, the analysis method review proposal unit 102 proposes review of the analysis example (step S205). For example, if the adoption rate of the K-means method exceeds the threshold in the clustering method, the registered / updated user of the analysis case not using the K-means method, the person in charge, the developer of the analysis method, or It is proposed that the analysis method is revised to the K-means method to providers and the like (hereinafter simply referred to as “users and the like”). Alternatively, if the adoption rate of the K-means method falls below the standard value in the clustering method, the analysis method is changed to a method different from the K-means method to users of analysis examples using the K-means method. Suggest to review. In this case, a list showing the analysis methods together with the adoption rate may be presented to the user or the like in descending order of the adoption rate.

<E-3. Effect>
In the data analysis method candidate determination device 16 according to the fifth embodiment, a review of the analysis method is proposed for an analysis example whose analysis purpose is the same as or similar to the analysis target data whose analysis method is determined by the analysis method candidate determination unit 4 An analysis method review proposal unit 102 is provided. In this way, by calculating the adoption rate for each analysis method in the past analysis cases and proposing a review of the analysis method based on the adoption rate, proposals for new analysis method candidates etc. can be made for past analysis cases. It is possible to improve the analysis accuracy when the analysis method is executed.

In the present invention, it is possible to freely combine the respective embodiments within the scope of the invention, and to appropriately modify and omit the respective embodiments.

Although the present invention has been described in detail, the above description is illustrative in all aspects, and the present invention is not limited thereto. It is understood that countless variations that are not illustrated can be envisaged without departing from the scope of the present invention.

2 analysis target data storage unit, 3 analysis case storage unit, 4 analysis method candidate determination unit, 5 input unit, 6 output unit, 7 evaluation acquisition unit, 8 recommended case storage unit, 9 attribute addition unit, 10 model change proposal unit, 11, 12, 13, 14, 15, 16 Data analysis method candidate determination device, 20 processor, 21 memory, 22 recording medium, 101 existing data utilization proposal unit, 102 analysis method review proposal unit.

Claims (13)

A data analysis method candidate determination device for determining an analysis method candidate of analysis target data to be analyzed,
An analysis case storage unit (3) for storing data associated with data attributes and analysis methods as analysis cases for each of a plurality of analyzed data that has been subjected to data analysis in the past,
About the analysis target data, an analysis target data storage unit (2) for storing data attribute information;
A data attribute similarity that is a similarity between the data attribute of the analysis target data and the data attribute of the analyzed data is calculated, and at least one analysis is performed from the analysis methods of the analyzed data based on the data attribute similarity An analysis method candidate determination unit (4) for determining a method as an analysis method candidate for the analysis target data,
Data analysis method candidate determination device. The analysis case storage unit (3) stores analysis purpose information for each of the plurality of analyzed data,
The analysis target data storage unit (2) stores analysis purpose information of the analysis target data,
The analysis method candidate determination unit (4) calculates the similarity between the analysis purpose of the analysis target data and the analysis purpose of the analyzed data as the analysis purpose similarity, and the analysis purpose similarity and the data attribute similarity The analysis target data and the analyzed data are calculated based on the total similarity, and based on the total similarity, at least one analysis method is selected from the analysis methods of the analyzed data. As determined,
The data analysis method candidate determination device according to claim 1. The data attribute of the analyzed data and the analysis target data includes at least one of a data acquisition interval, a data acquisition method, a result value, a predicted value, or a processed value,
The data analysis technique candidate determination apparatus according to claim 1 or 2. The analysis technique candidate determination unit (4) calculates the analysis target similarity based on the analysis target character string of the analysis target data and the analysis target character string of the analyzed data.
The data analysis method candidate determination device according to claim 2. The analysis technique candidate determination unit (4) calculates the analysis object similarity based on the analysis purpose of the analysis target data described in the hierarchical structure and the analysis purpose of the analyzed data described in the hierarchical structure.
The data analysis method candidate determination device according to claim 2. When the analysis purpose of the analysis target data and the analysis purpose of the analyzed data are described in source code or intermediate code,
The analysis technique candidate determination unit (4) indicates the processing procedure shown in the source code or the intermediate code for analysis of the analysis target data and the source code or the intermediate code for the analysis of the analyzed data. The data analysis technique candidate determination device according to claim 2, wherein the similarity to the processed processing procedure is calculated as the analysis target similarity based on a matching rate or continuity of matching processing procedures. The analysis method candidate determination unit (4) calculates, for each analysis method, an average value of the total similarity between the analyzed data using the analysis method and the analysis target data, and calculates the average of the total similarity The analysis method selected based on the value is determined as the analysis method candidate.
The data analysis method candidate determination device according to any one of claims 2, 4 to 6. An evaluation acquisition unit (7) for acquiring user evaluation information for the analysis technique candidate;
A recommended case storage unit (8) for storing data associating data attributes of the analysis target data, the analysis method candidates of the analysis target data, and the evaluation information for the analysis method candidates as recommended cases;
Further comprising
The data analysis technique candidate determination apparatus according to any one of claims 1 to 7. An attribute adding unit (9) for extracting the reason for non-adoption of the analysis technique candidate from the evaluation information acquired by the evaluation acquiring unit (7) and adding an item corresponding to the reason for non-adopting to the item of the data attribute In addition,
The data analysis method candidate determination device according to claim 8. The analysis case storage unit (3) stores information on the physical model before the change for the analysis case in which the data analysis is performed using the physical model in which the user has changed the physical model,
If the analysis method candidate is an analysis method using a physical model, and the physical model used in the analysis method candidate is the same as the physical model before the change in the analysis example, a model for proposing the change of the physical model A change proposal unit (10);
The data analysis method candidate determination device according to any one of claims 1 to 9. When the first analysis target data does not have data attributes necessary for the analysis method determined by the analysis method candidate determination unit (4) for the first analysis target data among the analysis target data. , Further comprising an existing data utilization proposing unit (101) that proposes to the user to utilize the second analysis object data having the necessary data attribute among the analysis object data,
The data analysis technique candidate determination apparatus according to any one of claims 1 to 10. The second analysis target data has a data attribute relating to whether or not the data can be used,
The existing data utilization proposing unit (101) provides the user with information on whether or not the second analysis target data can be used when proposing the user to utilize the second analysis target data.
The data analysis method candidate determination device according to claim 11. An analysis method review proposal unit (102) that proposes a review of the analysis method for the analysis example whose analysis purpose is the same as or similar to the analysis target data for which the analysis method candidate is determined by the analysis method candidate determination unit (4) In addition,
The data analysis method candidate determination device according to any one of claims 1 to 12.

Images