CN110866696A - Method and device for training shop falling risk assessment model - Google Patents

Abstract

The embodiment of the invention provides a method and a device for training a shop falling risk assessment model, wherein the method comprises the following steps: acquiring a training sample set, wherein the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data; training an initial model based on the training samples to obtain a target model; the objective model is used for estimating the risk of the shop falling. The model obtained through training can directly evaluate the shop falling probability, and the evaluation efficiency and the evaluation precision of shop falling are improved.

Description

Method and device for training shop falling risk assessment model

Technical Field

The application relates to the technical field of risk early warning, in particular to a training method and device for a shop falling risk assessment model.

Background

In recent years, under the situation of stable economic operation and continuous upgrading of consumption, the overall development trend of commercial property is still good. However, in the operation of commercial real estate, many risks and problems are faced, such as withdrawal of leases from merchants, adjustment of the state of business, and management of property. Among these, the risk of "shop dropping" (merchant quitting business) management is a major risk faced by many commercial property enterprises. The occurrence of 'shop dropping' not only reduces rent income, but also increases maintenance cost and re-enrollment cost, causing huge loss to enterprises. If risk monitoring and early warning can be carried out on 'fallen shops', the shops are helped in time to avoid the shops from being withdrawn from lease due to difficult operation, or follow-up recruitment work is done in advance, the loss caused by the merchants withdrawing from operation can be reduced, and the enterprise competitiveness is enhanced.

At present, the traditional shop floor-dropping analysis method mainly adopts manual analysis and prediction, and often needs to go through a plurality of steps of data collection, data arrangement, data analysis, result comparison, report writing and the like, thereby greatly occupying the working time of marketing personnel and ensuring that the working efficiency is lower. After modern businesses are added into the big data era, the defects of the traditional method are more and more obvious, the difficulty of extracting useful information from massive, high-speed and low-value-density data by using the traditional manual analysis method is extremely high, the accuracy and the timeliness are not high, and the monitoring of macroscopic risk conditions cannot be carried out.

Disclosure of Invention

According to the method and the device for training the shop falling risk assessment model, the shop falling risk early warning is performed by using the model obtained through training, and the problem that the traditional manual analysis efficiency is low is solved.

In order to improve the accuracy and efficiency of the shop floor estimation, the embodiment of the invention provides the following technical scheme:

the embodiment of the invention provides a training method for a shop falling risk assessment model, which comprises the following steps: acquiring a training sample set, wherein the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data; training an initial model based on the training samples to obtain a target model; the objective model is used for estimating the risk of the shop falling.

Optionally, the multiple indexes are divided into multiple index dimensions, where each index dimension corresponds to a part of the multiple indexes; the initial model is trained based on a plurality of metrics in a plurality of metric dimensions.

Optionally, the step of training the initial model based on the training samples to obtain the target model includes:

dividing time intervals, wherein one time interval corresponds to one sub-model; training according to the same steps to obtain each sub-model based on the training samples; and integrating all the sub-models obtained by training to obtain a final target model.

The application also provides a risk assessment model training device that shops fall, it includes: the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring training samples, the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data; a training module for training an initial model based on the training samples to obtain a target model; the objective model is used for estimating the risk of the shop falling.

Compared with the prior art, the invention has the beneficial effects that:

1. the model is trained on the basis of the multiple indexes through the multiple obtained indexes of the normally-running shops and the shops which are already subjected to shop dropping, the trained model can directly evaluate the shop dropping probability of the shops, and the evaluation efficiency and the evaluation accuracy of the shop dropping are improved.

2. The multiple indexes are divided into multiple index dimensions with pertinence, the model is trained based on the indexes in the multiple index dimensions, and the influence of the multiple index dimensions on the shop floor can be obtained from the evaluation result of the trained model on the shop floor, so that the shop can improve the operation strategy according to the evaluation result, and the shop floor risk is reduced.

3. Before the training sample is input into the model for training, all indexes in the training sample are subjected to distinguishing capability processing in a WOE (word on edge) binning mode, IV values of all the obtained indexes are compared, and only the index with strong distinguishing capability is reserved, so that the training efficiency and the prediction accuracy of the model are improved.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.

Fig. 1 is an electronic device according to an embodiment of the present disclosure;

FIG. 2 is a flowchart of a training method for a shop drop risk assessment model according to an embodiment of the present disclosure;

FIG. 3 is a flowchart illustrating the sub-steps of step S120;

FIG. 4 is a flowchart illustrating the sub-steps of step S122;

FIG. 5 is a schematic flow chart of the step of screening the index;

FIG. 6 is a block diagram of a training apparatus for a shop drop risk assessment model provided in the present application;

fig. 7 is a schematic structural diagram of an object model according to an embodiment of the present application.

FIG. 8 is a graph showing ROC curves in the experimental examples.

FIG. 9 is a PR graph in an experimental example.

Icon: 10-an electronic device; 12-a processor; 14-a memory; 100-a shop falling risk assessment model training device; 110-an obtaining module; 120-a training module; 130-inspection module.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.

The invention is realized by the following technical scheme.

The technical solution of the present invention will be described in detail below with reference to the accompanying drawings.

Referring to fig. 1, an electronic device 10 provided in an embodiment of the present application includes a memory 14 and a processor 12, where the memory and the processor are electrically connected directly or indirectly to implement data transmission or interaction, where the electronic device may be a server, a terminal device, or any device with data storage and processing capabilities.

The memory stores software functional modules stored in the memory in the form of software or firmware (firmware), and the processor executes various functional applications and data processing by running software programs and modules stored in the memory, such as the training device 100 for the shop floor risk assessment model in the embodiment of the present invention, so as to implement the training method for the shop floor risk assessment model in the embodiment of the present invention.

Referring to fig. 2, fig. 2 is a schematic flow chart of a training method for a shop floor risk assessment model according to an embodiment of the present disclosure, where the electronic device executes steps S110 to S130 when implementing the training method for the shop floor risk assessment model.

Step S110, a training sample is obtained.

In the embodiment of the present application, training samples may be obtained from a database, where the training samples include a good sample (a shop in normal operation) and a bad sample (a shop where a shop drop phenomenon has occurred), and a plurality of index data included in the good sample and the bad sample.

In the embodiment of the present application, the plurality of indicators may include, but are not limited to: first-class business, merchant property, contract days, berth type, agency mode, average increase rate of turnover, average change rate of turnover, average increase rate of passenger flow, maximum withdrawal of passenger flow, media activity, business area ratio of shops in the same square and the same industry, city type, business volume ratio of shops in the same square and the same industry.

In this embodiment, optionally, based on the database owned data, in combination with the business logic and various commonly used index construction techniques, the multiple indexes are divided into multiple index dimensions, where each index dimension corresponds to a part of the multiple indexes. For example: dividing a plurality of indexes into store traits, operation conditions, activity degree and competitive factors, wherein the corresponding part of indexes in the store characteristics can be first-level business state, merchant properties, contract days, berth types and agency modes; the corresponding part of the indexes in the operation condition can be the average turnover increase rate and the average turnover change value; part of indexes corresponding to the activity degree can be average passenger flow rate, maximum passenger flow withdrawal and media activity; the corresponding part of indexes in the competitive factors can be the business area ratio of the same-square and same-industry shop, the city type and the business turnover ratio of the same-square and same-industry shop.

The ① store speciality includes the relationship between store and state, the probability of the restaurant industry falling off is greater than the probability of the living boutique falling off and the probability of the clothing industry falling off, the relationship between the area of the store and the falling off, when the area of the store is between 100 square meters and 150 square meters, the store presents a lower falling off possibility, when the area of the store is greater than 330 square meters, the store may terminate the contract in advance along with the deterioration of the operating condition, the contract days and the falling off relationship, and when the contract year is about three, the probability of the falling off is relatively high.

②, the relationship between the turnover increase rate and the fallen shop, when the average turnover ratio per month is reduced by more than 5%, the probability of the fallen shop is increased, the relationship between the maximum withdrawal of the turnover and the fallen shop is increased, the greater the maximum withdrawal of the turnover in the last half year is, the relationship between the continuous decline of the turnover and the fallen shop is increased, if the turnover continues to decline for three months, the probability of the fallen shop is increased.

③, the activity degree is that the relationship between the passenger flow rate and the drop pavement, the monotonous linear relationship between the monthly average passenger flow rate and the drop pavement, the higher the monthly average passenger flow rate, the lower the drop pavement probability, the relationship between the maximum withdrawal of the passenger flow rate and the drop pavement, the monotonous linear relationship between the maximum withdrawal of the passenger flow rate and the drop pavement, the higher the drop pavement probability in the last half year, the relationship between the passenger flow rate and the drop pavement, the monotonous linear relationship between the monthly average passenger flow rate and the drop pavement, the higher the monthly average passenger flow rate, the lower the drop pavement probability, the relationship between the continuous decline of the passenger flow rate and the drop pavement, if the passenger flow rate of the shop continuously decreases for three months, the probability of the drop pavement of the shop becomes greater, and the initial model is trained based on a plurality of indexes in a plurality of index dimensions.

It is easy to understand that the selection of the index and the division of the dimension are only an illustrative example, and different indexes or more indexes can be adopted based on different considerations, and different dimension division can also be performed, and the method of the present invention has no hard requirement for this.

In the embodiment of the present application, obtaining samples further includes a sample sampling manner, and since bad samples (tiled samples) are not small in number and data provided in a database is already average sampled and spread across various cities and squares, when a training set test set is divided for existing data, a method of random division is directly adopted in the embodiment of the present application:

in the present example, 90% of all samples are used as training data sets and the remaining 10% are used as test data sets (which may be re-scaled if new data is available). In order to ensure the accuracy and stability of the model and reduce coefficient deviation, a Bootstrap sampling method is adopted to generate 100 groups of training samples according to the ratio of 1:4 of good and bad samples, each training sample is used for training an initial model respectively to generate 100 groups of coefficients, and each coefficient is averaged to obtain the final output value (coefficient) of the model.

Step S120, training the initial model based on the training sample to obtain the target model.

In the embodiment of the application, the target model is obtained after the initial model is trained for a plurality of times, and the target model is used for evaluating the risk of the shop falling. And training the initial model based on a plurality of indexes of a plurality of index dimensions in the training sample, and further obtaining a target model.

Step S130, the obtained target model is checked.

In the embodiment of the application, after the initial model is trained based on the multiple indexes of the multiple index dimensions in the training sample and the target model is obtained, in order to ensure the evaluation effect of the model, the trained target model can be tested. For example, the obtained integral model is verified on a reserved test data set, and the reliability and the accuracy of the model are verified by drawing an ROC curve and a PR curve and calculating the classification AUC, the classification accuracy and the classification recall ratio. As shown in fig. 8-9, the target model is analyzed by using the training sample data and the test sample data, which both show better effects, that is, the target model obtained by the method of the present invention has higher prediction ability for the risk of shop falling. Note that this step is an optional step, and is not a necessary step.

Referring to fig. 3, fig. 3 is a flowchart illustrating the sub-steps of step S120. The initial model described in this embodiment includes a plurality of submodels, where each submodel corresponds to a time interval. Specifically, the method comprises the following steps:

step S121, time intervals are divided, and one time interval corresponds to one sub-model.

In this embodiment, the current month in which the shop-out behavior occurs or the contract end of the normal operation is performed is used as a cut-off point, for example, the current month is respectively reversed for 1 month, 2 months, 3 months, 4 months, 5 months, and 6 months, and the training of the sub-model is performed based on the indicators of different time sections of the shop operation.

It is easy to understand that the time interval division and the number setting of the submodels are only an illustrative example, and there may be other different processing manners, such as one time interval every two months, or for example, a time interval of one year and one month is pushed forward, and then 12 submodels are obtained.

And S122, training according to the same steps to obtain each sub-model based on the training samples.

Referring to fig. 4, each sub-model training step includes:

in step S1221, a training sample is input.

In the embodiment of the invention, the training samples are input into the submodel, wherein a plurality of indexes in the training samples are indexes in a time interval corresponding to the submodel. For example, if 6 models correspond to one month and two months … …, and one model corresponds to one month before the time node, all the indexes in the training sample should be the indexes laid by the shop in the month. In case a submodel corresponds to a time interval of two months, all the indicators in the training sample should be those laid out in the store within the two months.

Step S1222, processing is performed based on the multiple indexes in the training sample, and evaluation values of multiple index dimensions are obtained.

In the embodiment of the present application, the evaluation values of the plurality of index dimensions are equivalent to the evaluation conditions of the store in the plurality of index dimensions, and the probability of the falling of the good sample and the bad sample can be obtained by the evaluation values.

Optionally, scoring is performed based on multiple indexes, and scoring values of multiple index dimensions are obtained. The scoring values of the multiple index dimensions are the sum of the scoring values of the multiple indexes. And carrying out weight distribution on the multiple index dimensions by combining the score of each index dimension to obtain the weights of the multiple index dimensions. The score value and the weight of each index dimension are taken as evaluation values.

In the embodiment of the application, a sub-model is trained by a logistic regression algorithm based on a plurality of indexes of a plurality of index dimensions in a training sample. For example, an index dimension is a store trait, and modeling is performed based primarily on a series of indices of the index dimension. The store traits include: the first-level business state, the second-level business state, the third-level business state, the logarithm of the business area, the special quality of the merchant, the contract days, the berth type and the agency mode total 8 indexes, the 8 indexes are used as variables to be input, and a scoring value which is related to the fact that the index dimension is the special quality of the shop is obtained and output according to the training of a logistic regression algorithm. The sub-model is thus trained on other index dimensions by a logistic regression algorithm.

In the embodiment of the application, each index dimension is subjected to weight distribution by adopting a genetic algorithm, and an objective function of the genetic algorithm is information entropy. And superposing the scoring values of all index dimensions based on the obtained weights, so as to obtain the submodel observed based on a certain period of time point.

In some embodiments, a sub-model is trained to obtain a score of each index dimension by dividing a plurality of indexes into the four index dimensions. And secondly, obtaining weights corresponding to the four index dimensions through a genetic algorithm, and obtaining a sub-model 1 after accumulation, namely a training model based on one month (time period corresponding to the sub-model) before the shop cut-off point. In order to obtain an optimal (local optimal) weight distribution in four index dimensions and ensure that the sum of the four weights is 1, a genetic algorithm is adopted for weight search. The method comprises the following specific steps: setting an optimization function of a genetic algorithm, wherein a cross entropy loss function is adopted, setting an initial population size, a maximum iteration number, a copy probability, a mutation probability and a cross probability, and finally, automatically stopping iteration after the optimization function is converged and returning the optimal result weight. For example, in one experimental example, the weights of the four index dimensions are 0.271, 0.239, 0.24, and 0.24, respectively.

Step S1223, the evaluation value obtained in step S1222 is checked, and if the error of the evaluation value reaches the set error precision or the iteration number reaches the set value, the training is stopped, otherwise, the method returns to step S1221, and new sample data is input for repeated training until the training is finished.

And S123, integrating all the sub-models obtained by training in the step S122 to obtain a final target model. In this embodiment, each submodel is assigned with a weight by using a genetic algorithm, and then the score values and the weights of the submodels are superimposed to obtain a final target model, i.e., a risk model capable of predicting the possibility of shop falling. The model outputs the final predicted probability of shop-dropping within half a year of the merchant, from 0 to 1, which indicates that the probability of shop-dropping is from low to high.

In step S110, after the determination of the index included in the sample is preliminarily determined according to the service logic and the prior knowledge, the distinguishing capability analysis may be performed to remove the index with poor distinguishing capability and retain the index with strong distinguishing capability.

Referring to fig. 5, fig. 5 is a schematic flow chart of the screening step for the index, and the specific screening process includes:

step S1101, performing WOE binning processing on the preliminarily determined indexes, and obtaining an IV value of each index.

In the embodiment of the application, before the training sample enters the model, WOE binning is performed on all index data in the training sample, variables with continuous values are divided into a plurality of discrete classes, one class is one binning, and variables with discrete values are divided into one binning, so that interference of extreme data, abnormal data and missing data on the model is avoided, and the training efficiency and the prediction accuracy of the model are improved. The IV value of a single variable can be measured based on WOE classification to evaluate the distinguishing capability of a single index for a shop.

In some embodiments, based on the existing data, 58 indexes are determined and divided into 4 large index dimensions (store speciality, business condition, activity degree and competitive factor), and most indexes show good prediction capability according to the IV value of each index. Before entering the model, WOE binning is firstly carried out on all field data of the test set, and variables with continuous values are divided into a plurality of discrete classes. And the discrete data is subjected to binning according to the number of the data types. Each type of discrete data will act as a box for the WOE. If the continuous data can be divided into five boxes, the continuous data is divided into five boxes; if the continuous variable cannot be divided into five boxes (in the case where a certain value appears a plurality of times), the changed value is set as a special value and divided into one box, and the remaining values are divided into five boxes. In the case where the above operation is still not in line, the first two values that appear more are set as special values, and divided into two separate bins, and the remaining data is divided into four bins.

Step S1102, determining the distinguishing capability of each index according to a preset IV value interval based on the IV value of the index.

In the embodiment of the application, based on the acquired IV value of each index, through a preset IV value interval, the index with weak distinguishing capability and strong distinguishing capability can be set, and the index with weak distinguishing capability, medium distinguishing capability and strong distinguishing capability can also be set, and the preset IV value can be set by a user on demand. The IV value represents the occupation difference of good and bad samples of the variable in different value groups, and the larger the IV value is, the larger the discrimination capability of the index on the good and bad samples is. The preset interval of IV values may be, for example: IV at [0.02,0.1), the index has weak discriminative power, IV at [0.1,0.3), the index has medium discriminative power, IV is greater than or equal to 0.3, the index has strong discriminative power. The index of weak discrimination is deleted and the indexes of medium discrimination and strong discrimination are retained. For example, the value of the media activity indicator IV is 0.000, which means that the indicator has almost no distinguishing capability, so the media activity indicator is screened out when entering the module; the merchant property index IV value is 0.334, which indicates that the index has strong distinguishing capability and keeps the merchant property index.

Referring to fig. 6, fig. 6 is a schematic structural diagram of a training device for a shop risk assessment model, where the training device includes an obtaining module, a training module, and a checking module.

The acquisition module is used for acquiring training samples, wherein the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data.

In the embodiment of the present application, the obtaining module is configured to execute step S110 in fig. 2, and reference may be made to the detailed description of step S110 in fig. 2 for the detailed description of the obtaining module.

The training module is used for training an initial model based on the training sample to obtain a target model; the objective model is used for estimating the risk of the shop falling.

In the embodiment of the present application, the training module is configured to perform step S120 in fig. 2, and the detailed description about the training module may refer to the detailed description about step S120 in fig. 2.

And the inspection module is used for inspecting the obtained target model.

In the embodiment of the present application, the checking module is configured to perform step S130 in fig. 2, and the detailed description about the checking module may refer to the detailed description about step S130 in fig. 2.

In some embodiments, the initial model comprises a plurality of sub-models, wherein each sub-model corresponds to a time interval; the training module comprises a dividing submodule, a training submodule and an integrating submodule.

The division submodule is used for dividing time intervals, and each time interval corresponds to one submodel. For example, the time interval division is performed by respectively reversing 1 month, 2 months, 3 months, 4 months, 5 months and 6 months forward by taking the current month of the collapse or the contract end of normal operation as a cut-off point.

In the embodiment of the present application, the dividing sub-module is configured to perform step S121 in fig. 3, and the detailed description about the checking module may refer to the detailed description about step S121 in fig. 3.

And the training submodule is used for training to obtain each sub-model according to the same steps based on the training sample.

In the embodiment of the present application, the input module is configured to execute step S122 in fig. 3, and the detailed description about the verification module may refer to the detailed description about step S122 in fig. 3.

And the integration submodule is used for integrating all the sub-models obtained by training to obtain a final target model.

In the embodiment of the present application, the input module is configured to execute step S123 in fig. 3, and the detailed description about the inspection module may refer to the detailed description about step S123 in fig. 3.

Referring to fig. 7, fig. 7 is a schematic structural diagram of a target model according to an embodiment of the present disclosure.

In the embodiment of the application, the store characteristics, the business condition, the activity degree and the competitive factors are targeted index dimensions divided by a plurality of indexes. Model 1, model 2, model 3, model 4, model 5, and model 6 are six submodels. The cutoff point is the date that the business operation contract expires, and the corresponding cutoff point is the time interval corresponding to each submodel.

Those of ordinary skill in the art will appreciate that the elements and algorithm steps of the examples described in connection with the embodiments disclosed herein may be embodied in electronic hardware, computer software, or combinations of both, and that the components and steps of the examples have been described in a functional general in the foregoing description for the purpose of illustrating clearly the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the implementation. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention essentially or partially contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (10)

1. A training method for a shop deckbay risk assessment model is characterized by comprising the following steps:

acquiring a training sample set, wherein the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data;

training an initial model based on the training samples to obtain a target model; the objective model is used for estimating the risk of the shop falling.

2. The method of claim 1, wherein the plurality of metrics are divided into a plurality of metric dimensions, wherein each metric dimension corresponds to a portion of the plurality of metrics;

the initial model is trained based on a plurality of metrics in a plurality of metric dimensions.

3. The method of claim 2, wherein the step of training an initial model based on the training samples to obtain a target model comprises:

dividing time intervals, wherein one time interval corresponds to one sub-model;

training according to the same steps to obtain each sub-model based on the training samples;

and integrating all the sub-models obtained by training to obtain a final target model.

4. The method of claim 3, wherein the step of training each sub-model based on the training samples according to the same steps comprises:

inputting a training sample, wherein a plurality of index data in the training sample are index data in a time interval corresponding to the sub-model;

performing logistic regression training based on the indexes in the training sample to obtain evaluation values of the multiple index dimensions;

and checking the obtained evaluation value, stopping training if the error of the evaluation value reaches the set error precision or the iteration frequency reaches a set value, or inputting new sample data to perform repeated training until the training is finished.

5. The method of claim 3, wherein the step of integrating all trained submodels comprises: and (4) carrying out weight distribution on all the sub-models through a genetic algorithm, and then superposing the score values and the weights of all the sub-models to obtain a final target model.

6. The method of claim 1, wherein the indicators in the sample are determined by:

performing WOE (word on average) binning processing on each preliminarily determined index, and obtaining an IV (input/output) value of each index;

and determining the distinguishing capability of each index according to a preset IV value interval, reserving the index with strong distinguishing capability, and removing the index with weak distinguishing capability, wherein the index in the sample is the reserved index.

7. The utility model provides a shop risk assessment model trainer that decks which characterized in that, it includes:

the system comprises an acquisition module, a processing module and a processing module, wherein the acquisition module is used for acquiring training samples, the training samples comprise good samples and bad samples, and each sample comprises a plurality of index data;

a training module for training an initial model based on the training samples to obtain a target model; the objective model is used for estimating the risk of the shop falling.

8. The apparatus of claim 7, wherein the plurality of metrics are divided into a plurality of metric dimensions, wherein each metric dimension corresponds to a portion of the plurality of metrics;

the initial model is trained based on a plurality of metrics in a plurality of metric dimensions.

9. The apparatus of claim 8, wherein the initial model comprises a plurality of submodels, wherein each submodel corresponds to a time interval; the training module comprises:

the division submodule is used for dividing time intervals, and each time interval corresponds to one submodel;

the training submodule is used for training according to the same steps to obtain each submodel based on the training sample;

and the integration submodule is used for integrating all the sub-models obtained by training to obtain a final target model.

10. A shop drop risk assessment model obtained by training with the training method of the shop drop risk assessment model according to any one of claims 1 to 6, wherein the shop drop risk assessment model is used for assessing the shop drop risk.

Images