WO2015129988A1 - Method for generating category preference for each user, method for product recommendation using same, and device for same - Google Patents

Abstract

Disclosed are a method for generating category preference for each user, a method for product recommendation, and a device for same. Category preference having a degree of interest of a user appropriately applied thereto is calculated by using a recency weight, products to be recommended for each category are selected by using the category preference, and exposure priority of the selected products is determined on the basis of tendency which corresponds to the development of changes in the frequency of use during a reference period for each category, thereby enabling more appropriate product recommendation for the user.

Description

Category preference method for each user, product recommendation method using the same, and apparatus therefor

The present invention relates to a method for generating category preference for each user, a method for recommending a product using the same, and an apparatus for the same, particularly the latest for each user. Creates service category preferences in consideration of gender, and recommends products / content / services suitable for the user in consideration of category preferences, but determines category preferences by user in consideration of category tendency. It relates to a production method, a product recommendation method using the same, and an apparatus therefor.

The present invention claims the benefit of the filing date of Korean Patent Application No. 10-2014-0022933, filed February 26, 2014, the entire contents of which are incorporated herein.

A user using a content providing service such as an online shopping mall, an app market, a movie, or the like has relatively preferred categories among the categories existing in the service.

That is, user A using an online shopping mall searches for or purchases a product in the home appliance-computer category in most cases visiting the online shopping mall, and user B uses a product in the clothing-bag category in most cases visiting the online shopping mall. Purchase or search.

Meanwhile, as online shopping malls, app markets, and content providing services are becoming more popular, products such as collaborative filtering and association rule mining are recommended to users in consideration of usage patterns of other users in a group similar to the user. Product recommendation technology is being researched. As such, when product recommendation is made in consideration of usage patterns of other users, category preference information of similar groups of users is roughly reflected, but category preferences of individual users are not accurately reflected, so it is limited to recommend products optimized for individual users. There is. That is, since one user uses the products of a plurality of categories in different distributions, if a general recommendation algorithm is applied without considering the user's category preference, the optimal product recommendation is difficult.

Accordingly, there is an urgent need for a new technology that more accurately reflects the degree of category preference of individual users and calculates the category preference, and can effectively use it for product (content, service) recommendation.

Related prior arts include Korean Patent Application Publication No. 10-2011-0071452 and June 29, 2011 (name: broadcast program recommendation method and apparatus using genre weight).

An object of the present invention is to calculate the category preference considering the recency for each user to index the recent category preference degree of the user.

In addition, an object of the present invention is to calculate the category preference that effectively reflects the freshness of the category preference degree using the freshness weight generated using the exponential function.

In addition, an object of the present invention is to select a product recommended to the user by using the category preference for each user to enable a more suitable product recommendation to the user.

In addition, an object of the present invention is to determine the exposure order for the recommended product using the category preference for each user in consideration of the tendency of the preference for each category to enable more appropriate product recommendation in accordance with the trend of user preference changes.

According to an aspect of the present invention, there is provided a category preference generating apparatus, including: a storage unit storing a number of times of use of each category during a reference period for each user for each of the unit periods; And a controller configured to calculate a preference for each category corresponding to the reference period by using recency weights for one or more of the unit periods and the number of times of use per category.

In this case, the controller may calculate the preference for each category corresponding to the reference period in which the latestness is reflected by summing the number of times of use of each category to which the latest weights are applied.

In this case, the controller applies a first recency weight to a first unit period among the unit periods, and a second latest value greater than the first recency weight in a second unit period closer to the current time point than the first unit period. Gender weights can be applied.

In this case, the freshness weight is calculated by applying an exponential function, and the freshness weight is adjusted by adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function. The degree of reduction of these can be controlled.

In this case, the controller may calculate a trend line corresponding to a change in the number of times of use during the reference period, and calculate a preference for each category corresponding to the reference period in which a trend is reflected according to the slope of the trend line.

In this case, the controller may control such that the trend weight is applied to a category having a positive value higher than a category having a negative value.

In addition, the product recommendation apparatus according to the present invention corresponds to the reference period by using the recency weights for any one or more of the unit periods and the number of times of use for each category during the reference period for each of the users. A category preference generator for calculating a preference for each category; A recommendation SET generator generating the recommendation set for each category; And recommending the recommended products by selecting the recommended products in the recommendation set for each category in consideration of one or more of the tendency corresponding to the change in the number of times of use during the reference period and the preference of each category. Contains groups.

In this case, the tendency may be calculated based on the slope of the trend line corresponding to the change trend of the use frequency.

In this case, the product recommender may preferentially recommend a product of a category in which the slope of the trend line has a positive value over a product of a category in which the slope of the trend line has a negative value.

In this case, the product recommender may include a recommendation product selector for selecting recommendation products in the recommendation set for each category; A tendency generator for calculating the tendency by calculating a change in the number of times of use for each category; A recommendation rank controller for determining recommendation priorities of the recommended products based on the tendency; And an output interface for providing the recommendation products to a user according to the recommendation priority.

In this case, the category preference generator may calculate the category-specific preference corresponding to the reference period in which the latestness is reflected by summing the number of use of each category to which the latest weights are applied for each category.

In this case, the category preference generator applies a first freshness weight to a first unit period among the unit periods, and includes a first greater than the first freshness weight in a second unit period closer to the current time point than the first unit period. 2, the preference for each category may be calculated by applying the freshness weight.

In this case, the freshness weight is calculated by applying an exponential function, and the freshness weight is adjusted by adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function. The degree of reduction of these can be controlled.

In addition, the method for generating category preferences for each user according to the present invention may include: storing the number of times of use of each category during each reference period for each user; Calculating freshness weights for any one or more of the unit periods; And calculating a preference for each category corresponding to the reference period by using the freshness weights and the number of times of use of each category.

In this case, the calculating of the freshness weights may include calculating a second most recent weight applied to a first unit period among the unit periods, and applying a second most recent weight to a second unit period closer to the current time point than the first unit period. The freshness weights may be calculated to be less than the gender weight.

In this case, the freshness weight is calculated by applying an exponential function, and the freshness weight is adjusted by adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function. The degree of reduction of these can be controlled.

In addition, the product recommendation method according to an embodiment of the present invention, generating a recommendation set for each category; Calculating category-specific preferences corresponding to the reference periods using recency weights for any one or more of the unit periods and the number of use of each category during the reference period for each of the users; Selecting recommended products in the recommendation set for each category according to the preference for each category; And recommending the recommended products according to a priority corresponding to a trend corresponding to a change in the number of times of use during the reference period.

In this case, the tendency may be calculated based on the slope of the trend line corresponding to the change trend of the use frequency.

In this case, the recommendation may be to preferentially recommend products of the category in which the slope of the trend line has a positive value over products of the category in which the trend line has a negative value.

According to the present invention, a category preference in consideration of the recency for each user may be calculated to index the recent category preference of the user.

In addition, the present invention can calculate the category preference that effectively reflects the freshness of the category preference degree using the freshness weight generated using the exponential function.

In addition, the present invention can select a product recommended to the user by using the category preference for each user it is possible to recommend a more suitable product to the user.

In addition, the present invention can determine the exposure order for the recommended product using the category preference for each user in consideration of the tendency of the preference for each category, it is possible to more appropriate product recommendation in accordance with the trend of changing preferences of the user.

1 is a block diagram showing a product recommendation apparatus according to an embodiment of the present invention.

2 is a graph showing the number of times of use during the reference period for each category.

3 is a table showing the freshness weights assigned per unit periods for the values of two parameters a.

4 is a graph illustrating an exponential weight application for category # 1.

5 is a graph showing the application of the exponential weight for category #M.

6 is a graph showing trends for each category.

FIG. 7 is a block diagram illustrating an example of a category preference generator shown in FIG. 1.

8 is a block diagram illustrating an example of the recommended SET generator illustrated in FIG. 1.

FIG. 9 is a block diagram illustrating a product recommender illustrated in FIG. 1.

10 is a flowchart illustrating a method of generating category preferences according to an embodiment of the present invention.

11 is a flowchart illustrating a product recommendation method according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, in the following description and the accompanying drawings, detailed descriptions of well-known functions or configurations that may obscure the subject matter of the present invention will be omitted. In addition, it should be noted that like elements are denoted by the same reference numerals as much as possible throughout the drawings.

The terms or words used in the specification and claims described below should not be construed as being limited to ordinary or dictionary meanings, and the inventors are appropriate as concepts of terms for explaining their own invention in the best way. It should be interpreted as meanings and concepts in accordance with the technical spirit of the present invention based on the principle that it can be defined. Therefore, the embodiments described in the present specification and the configuration shown in the drawings are only the most preferred embodiments of the present invention, and do not represent all of the technical ideas of the present invention, and various alternatives may be substituted at the time of the present application. It should be understood that there may be equivalents and variations. In addition, terms such as “first” and “second” are used to describe various components, and are only used to distinguish one component from another component and are not used to limit the components.

1 is a block diagram showing a product recommendation apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the product recommendation apparatus 100 according to an exemplary embodiment includes a category preference generator 110, a product recommender 120, and a recommendation SET generator 130.

The category preference generator 110 uses the recency weights for any one or more of the unit periods and the number of use of each category during the reference period for each of the users, and the category preferences corresponding to the reference period. To calculate.

That is, the category preference generator 110 generates category preferences of individual users for a plurality of users who use a particular service. In this case, the category preference generator 110 may generate the category preference of the individual user in consideration of the latestness and tendency of the service usage history of the individual user.

The number of categories for products to be included in the recommended product may be determined based on the category preference calculated through the category preference generator 110. For example, if you have 36%, 34%, and 30% preference ratios for categories A, B, and C, and the total number of items recommended is 100, there are 36 items in category A and 34 items in category B. , 30 items of category C may be selected.

In this case, the category preference generator 110 may calculate the category-specific preference corresponding to the reference period in which the latestness is reflected by summing the number of times of use of the category to which the latest weights are applied for each category.

That is, the category preference generator 110 may consider the most recent number of times of use of each category more importantly in generating the category preference. As such, the preference of the user may be more appropriately reflected by calculating the category preference by reflecting the user's recent preference of preference.

In this case, the category preference generator 110 applies a first freshness weight to a first unit period among the unit periods, and the first freshness weight to a second unit period closer to the current time point than the first unit period. The preference for each category may be calculated by applying a larger second freshness weight.

In this case, the freshness weight is calculated by applying an exponential function, and by adjusting the absolute value of the value obtained by dividing the exponential function with respect to a period and dividing the exponential function by the exponential function. The degree of reduction of the weights can be controlled.

In the case of applying a linearly changing weight without using an index function, it is difficult to properly reflect the freshness of the user's interest in purchasing a product.

The recommendation SET generator 130 generates a recommendation set for each category.

That is, the recommendation SET generator 130 selects recommended products for each category and generates a recommendation set which is a set of recommended products.

The product recommender 120 selects and recommends the selected products in the recommendation set for each category in consideration of one or more of the tendency corresponding to the change in the number of times of use during the reference period and the preference for each category. Recommend products.

At this time, the product recommender 120 may select the recommended products in the recommendation set in consideration of the preference for each category, and determine the exposure priority of the selected recommended products in consideration of the tendency.

At this time, the product may be a concept encompassing all types of objects that can be traded, such as goods, services, content.

For example, the product recommender 120 may recommend goods traded through an online commerce site, recommend content such as an application or e-book traded through an app store, or recommend a service. have.

In this case, the tendency may be calculated based on the slope of the trend line corresponding to the change trend of the use frequency.

In this case, the tendency may be calculated based on the data before applying the exponential weight, or may be calculated based on the data after applying the exponential weight. In particular, when the tendency is calculated based on the data after applying the index weight, it is possible to calculate the tendency based on the preference weight reflecting the latestness.

In this case, the product recommender 120 may preferentially recommend a product of a category in which the slope of the trend line has a positive value over a product of a category in which the slope of the trend line has a negative value. .

2 is a graph showing the number of times of use during the reference period for each category.

Referring to FIG. 2, it can be seen that the number of times a user uses the corresponding category for each category period is accumulated for the category # 1, the category # 2, and the category #M (M is a natural number larger than 2).

In this case, the reference period (reference period # 1) may be a period in which the number of times of use of the user is recorded for each category. At this time, the reference period may be determined differently according to the service. For example, a service that targets content is sold on a daily basis, a service that sells a physical product on a weekly basis, or as a book. In the case of a service targeting a product having a long purchase cycle, a reference period may be set in months.

The reference period may be determined by how often users of the service 'use' the product.

In the example shown in FIG. 2, the reference period (reference period # 1) has eight unit periods (T-8, T-7, T-6, T-5, T-4, T-3, T-2 , T-1). The unit period T-1 is the most current time point among the eight unit periods T-8, T-7, T-6, T-5, T-4, T-3, T-2, and T-1. It can be a unit period close to.

In this case, the term 'use' means that the user is provided with a specific service through the corresponding product. For example, 'use' may mean that the user 'purchases' a specific product through the online market. For example, 'use' may mean that the user 'clicks' a specific application through the app store.

Thus, the number of times of use may be the number of times the user purchases the product in the online market. In addition, the number of times of use may be the number of times the user clicks the product in the App Store.

At this time, the unit period is a period in which the number of times of use is calculated, such as T-8, T-7, T-6, ..., and T-1 shown in FIG. For example, the unit period may be a period of one day, one hour, one week, or the like, and may be determined according to the type of service or product.

The preference for each category may be calculated by summing the number of times of usage per unit period during the reference period shown in FIG. 2.

For example, as shown in FIG. 2, for a service in which M categories exist, the number of times a user uses category # 1 for a reference period (reference period # 1) is 18 times, and the reference period (reference period # 1) the user can use the category # 2 for 17 times, and the user can use the category #M for 15 times during the reference period (reference period # 1). Since there are 50 total uses (18 + 17 + 15), assuming no other category other than category # 1, category # 2, and category #M, this user's category preference is 36% for category # 1, category 34% for 2 and 30% for category #M.

As such, the preference of each category of the user may be determined according to the ratio of the number of times the user uses the corresponding category.

However, if the user's preference for each category is calculated based on the ratio of the number of times the user uses the corresponding category, the category preference may change greatly when the reference period is changed. When the reference period # 2 is applied instead of the reference period # 1 illustrated in FIG. 2, the category preference of the user of the simple counting method is significantly changed.

In addition, as in the category (M) shown in Figure 2 used in the past (T-8 ~ T-5), but not recently used (T-4 ~ T-1) in the category, the category is used recently However, due to past use times, category preference may be high. In other words, if the reference period is long, a category that is rarely used in recent years may remain high for a certain period of time due to a history of suddenly used one or two times, which may be an obstacle to calculating accurate category preference of the user. have.

In particular, since the user's interest continuously changes over time, the recency of the preference needs to be reflected in the category preference calculation.

For example, in the App Market where apps that fall into categories such as Game / Life / Language / Movie / Ebook / Manga are sold, the user is mainly the e-book category in period A, the comics category in period B, and the comic category in the period Game categories are available. Furthermore, other users may continue to use game and language categories.

Therefore, it is difficult to reflect the recent user's interest in simply accumulating the number of times the user uses the corresponding category during the reference period.

In order to reflect the recent interest of the user, a weight may be applied to the number of times of use corresponding to the recent unit period among the unit periods. That is, when a greater weight of recency weight is assigned to unit periods close to the current time point among the unit periods included in the reference period, category preferences reflecting recent user interests may be obtained.

In the example shown in FIG. 2, as the weight (latest weight) corresponding to each unit period decreases from the unit period T-1 to the unit period T-8, the number of times of use close to the present time is more important. This is reflected in the category preference calculation.

For example, the freshness weight may change linearly. That is, the freshness weight may be linearly reduced from the current time point to the previous time point. For example, based on the current time T, the unit weight T1 is 1.0 for freshness, T-2 is 0.9 for freshness, T-3 is 0.8 for freshness, and TP is 1.0-0.1 * (P -1) can be assigned. In this case, 0.1 may be viewed as a step value.

For example, in the case of category # 1 shown in FIG. 2, the number of times of use from the unit period T-1 to T-8 was (2, 0, 3, 2, 3, 0, 3, 5) above 0.1. In the case of applying the decreasing linear weight as the freshness weight, it may be (2, 0, 3 * 0.8, 2 * 0.7, 3 * 0.6, 0, 3 * 0.4, 5 * 0.3).

When the freshness weight is linearly reduced, it is difficult to actively reflect the freshness because the weight is reduced by a certain slope for the entire reference period. In addition, it is necessary to consider the entire reference period when specifying a step value whose weight changes. For example, if the reference period includes 12 unit periods, and the weight change step value is 0.1, the weight for the unit period (T-12) has a value less than 0, and thus appropriate weight for all unit periods. In order to allocate, the number of unit periods included in the reference period should be considered when determining the step value.

Of course, the linear weight variation may be effective according to the type of service.

The freshness weight may decrease exponentially from the unit period close to the current point to the past. That is, the exponential weight may be applied as the freshness weight, which is calculated using an exponential function.

For example, when the current time is T, the freshness weight for the unit period T-P may be set to EXP (-a * (P-1)).

In this case, EXP () is an exponential function and a is a parameter for adjusting the weight reduction degree.

For example, the freshness weight for the unit period (T-1) (if P = 1) would be EXP (-a * 0) = 1 and gradually applied in the past, such as T-2 and T-3. The freshness weight will be reduced exponentially.

By applying the freshness weight calculated using the exponential function in this way, it is possible to determine the parameter a considering the degree of weight reduction without considering how many unit periods the reference period includes.

3 is a table showing the freshness weights assigned per unit periods for the values of two parameters a.

Referring to FIG. 3, if the user's recent interest is to be greatly reflected in the category preference calculation, the parameter a is set large (a = 0.4), the recent preference is reflected, but the weight is gradually decreased for the unit intervals. If you want, you can set the parameter a small (a = 0.1).

In this way, if the parameter a is appropriately set according to the service type or the like, the category preference reflecting the user's recent interest level can be calculated.

4 is a graph illustrating an exponential weight application for category # 1.

Referring to FIG. 4, in the graph after the freshness weight is applied, the number of uses corresponding to the unit period close to the present time is more important than the number of uses corresponding to the past unit period in the graph after the freshness weight is applied. It can be seen that.

5 is a graph showing the application of the exponential weight for category #M.

Similarly, referring to FIG. 5, it can be seen that the number of uses corresponding to the unit period close to the current time point is considered more important than before the freshness weight is applied.

As such, the category preference may be calculated in consideration of the latest weights to properly reflect the user's recent interest.

The number of times of use by category of the user may have a tendency within the reference period.

According to an embodiment, the freshness weight may reflect the tendency corresponding to the slope of the trend line corresponding to the change in the number of times of use during the reference period.

That is, if the number of times of use of the user increases within the reference period, a high tendency weight may be applied to the category preference of the corresponding category. On the contrary, if the number of times of use of the user decreases within the reference period, a low tendency weight may be applied to the category preference of the corresponding category.

Furthermore, the tendency corresponding to the change in the number of times of use during the reference period is not used to calculate category preference, but may be used when determining the priority of exposing products selected as recommended products to the user.

6 is a graph showing trends for each category.

Referring to FIG. 6, it can be seen that the number of times of use of the category # 1, the category # 2, and the category # 3 is equal to 16 times during the reference period.

In this case, although the above-described freshness weight may be applied for category preference calculation or product recommendation, trend weighting may be applied.

That is, it is difficult to determine which category of products to give priority to recommendation to the user because the category # 1, # 2 and category # 3 shown in FIG. Of course, in the case of applying the freshness weight, the freshness weight plays a role of adjusting the preference weight for each category, and thus the products of the category with high category preference may be preferentially exposed. However, it is difficult to reflect the user's taste sufficiently only with the latest.

As shown in FIG. 6, in the category # 1, the trend of changing the number of use of the user is decreasing, in the category # 2, the trend of increasing the number of usage of the user is increasing, and in the case of category 3, the trend of changing the number of usage of the user Is a maintaining trend. As such, if the trend corresponding to the change in the number of times of use during the reference period is reflected, more suitable products can be recommended to the user.

In this case, the trend corresponding to the change in the number of times of use during the reference period may be calculated based on the slope of the trend line corresponding to the change in the number of times of use during the reference period.

In this case, the tendency may be calculated based on the data before applying the exponential weight of FIG. 4, or may be calculated based on the data after applying the exponential weight of FIG. 4. In particular, when the tendency is calculated based on the data after applying the exponential weight of FIG. 4, it is possible to calculate the tendency based on the preference weight in which the latestness is reflected.

For example, a product of category # 2 of FIG. 6 having a positive slope of the trend line may be preferentially exposed to a product of category # 1 of FIG. 6 having a negative slope of the trend line. In this case, when the slope of the trend line is positive for both categories, a product corresponding to a category having a larger slope of the trend line may be preferentially recommended.

FIG. 7 is a block diagram illustrating an example of a category preference generator shown in FIG. 1.

Referring to FIG. 7, the category preference generator illustrated in FIG. 1 includes a storage unit 710, a controller 720, and a communication unit 730.

In this case, the category preference generator may correspond to the category preference generating apparatus described in the claims.

The storage unit 710 stores the number of times of use of each category during the reference period for each user for each of the unit periods.

The controller 720 calculates the category-specific preference corresponding to the reference period by using the recency weights for the at least one of the unit periods and the number of times of use per category.

In this case, the controller 720 may calculate the preference for each category corresponding to the reference period in which the latestness is reflected by summing the number of times of use of each category to which the latest weights are applied.

At this time, the controller 720 applies the first recency weight to the first unit period among the unit periods, and the second latest value greater than the first recency weight to the second unit period closer to the current time point than the first unit period. Gender weights can be applied.

In this case, the freshness weight is calculated by applying an exponential function, and the degree of reduction of the freshness weights can be controlled by adjusting the absolute value (a) of the value obtained by dividing the exponential function by the exponential function.

According to an embodiment, the controller 720 may calculate a trend line corresponding to a change in the number of times of use during the reference period, and calculate a preference for each category corresponding to the reference period in which the trend is reflected according to the slope of the trend line. In this case, the controller 720 may control a trend weight higher than a category having a negative value to a category having a positive value of the trend line.

The communication unit 730 transmits and receives information related to a plurality of terminals through a communication network such as a network. In particular, the communication unit 730 according to an embodiment of the present invention receives a request for a service from a terminal and provides an execution result corresponding to the service requested by the terminal to the terminal.

The storage unit 710 stores various information generated in the service process according to an embodiment of the present invention, in addition to the number of times of use of each category.

According to an embodiment, the storage unit 710 may be configured independently of the category preference generator 110 to support a function for a service. In this case, the storage unit 710 may operate as a separate mass storage, and may include a control function for performing an operation.

In addition, the category preference generator 110 configured as described above may be implemented in one or more hardware.

Meanwhile, the category preference generator 110 may include a memory to store information in the device. In one embodiment, the memory is a computer readable medium. In one implementation, the memory may be a volatile memory unit, and for other implementations, the memory may be a nonvolatile memory unit. In one embodiment, the storage device is a computer readable medium. In various different implementations, the storage device may include, for example, a hard disk device, an optical disk device, or some other mass storage device.

8 is a block diagram illustrating an example of the recommended SET generator illustrated in FIG. 1.

Referring to FIG. 8, the recommendation SET generator illustrated in FIG. 1 includes a user data collector 810, a category # 1 recommendation SET generator 820-1, a category # 2 recommendation SET generator 820-2, ..., Category #M recommendation SET generator 820-M.

The user data collector 810 collects user data such as a purchase history or a click history of a user so that the user data collector 810 may be used to generate a category recommendation set.

The category # 1 recommendation SET generator 820-1 generates a category # 1 recommendation group by selecting all or part of products belonging to category # 1 in consideration of user data.

Similarly, the category # 2 recommendation SET generator 820-2 generates a category # 2 recommendation group by selecting all or part of products belonging to category # 2 in consideration of user data.

Similarly, the category #M recommendation SET generator 820-M selects all or some of the products belonging to the category #M in consideration of user data to generate the category #M recommendation group.

At this time, the size of the recommended group for each category (the number of products belonging to the group) may all be the same, or may vary depending on the category.

FIG. 9 is a block diagram illustrating a product recommender illustrated in FIG. 1.

Referring to FIG. 9, the product recommender 120 illustrated in FIG. 1 includes a tendency generator 910, a recommendation product selector 920, a recommendation rank adjuster 930, and an output interface 940.

The recommendation product selector 920 selects the recommended products in the recommendation set for each category in consideration of the preferences for each category.

The tendency generator 910 generates a tendency by calculating a change trend of the number of times of use for each category.

In this case, the trend may be calculated based on the slope of the trend line corresponding to the change in the number of times of use within the reference period.

The recommendation ranking controller 930 determines recommendation priorities of recommendation products based on the trend.

At this time, the recommendation ranking controller 930 is a category of the product of the category in which the trend line has a negative value of the slope of the trend line corresponding to the change in the number of times of use within the reference period. We can recommend preferentially over product.

The output interface 940 exposes the recommended products to the user according to the recommendation priority.

10 is a flowchart illustrating a method of generating category preferences according to an embodiment of the present invention.

Referring to FIG. 10, in the method of generating category preferences according to an embodiment of the present invention, the number of times of use of each category during the reference period for each user is stored for each unit period (S1010).

In addition, the category preference generation method according to an embodiment of the present invention calculates freshness weights for any one or more of the unit periods (S1020).

In this case, in operation S1020, a first recency weight applied to a first unit period among the unit periods is greater than a second recency weight applied to a second unit period closer to a current time point than the first unit period. The freshness weights can be calculated to be small.

In this case, the freshness weight is calculated by applying an exponential function, and the freshness weight is adjusted by adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function. The degree of reduction of these can be controlled.

In addition, the category preference generation method according to an embodiment of the present invention calculates the category preferences corresponding to the reference period by using the freshness weights and the number of times of use per category (S1030).

11 is a flowchart illustrating a product recommendation method according to an embodiment of the present invention.

Referring to FIG. 11, the product recommendation method according to an embodiment of the present invention generates a recommendation set for each category (S1100).

In addition, the product recommendation method according to an embodiment of the present invention is based on the criteria using the recency weights for any one or more of the unit periods and the number of times of use of each category during the reference period for each user The preference for each category corresponding to the period is calculated (S1120).

In addition, the product recommendation method according to an embodiment of the present invention selects recommended products within the recommendation set for each category according to the category preferences (S1130).

In addition, the product recommendation method according to an embodiment of the present invention recommends the recommended products according to the priority corresponding to the tendency corresponding to the change in the number of times of use during the reference period (S1140).

In this case, the tendency may be calculated based on the slope of the trend line corresponding to the change trend of the use frequency.

In this case, step S1140 may preferentially recommend a product of a category in which the trend line has a positive value over a product of a category in which the trend line has a negative value.

The steps shown in FIGS. 10 and 11 may be performed in the order shown in FIGS. 10 and 11, in the reverse order, or simultaneously.

The category preference generation method and the product recommendation method according to the present invention may be implemented as a program or a smartphone app that can be executed through various computer means. At this time, the program or the smartphone app may be recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the media may be those specially designed and constructed for the purposes of the present invention, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, such as floppy disks. Magneto-optical media, and any type of hardware device specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions may include high-level language code that can be executed by a computer using an interpreter as well as machine code such as produced by a compiler. Such hardware devices may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

As described above, the method for generating category preferences for each user and the product recommendation method according to the present invention may not be limitedly applied to the configuration and method of the embodiments described as described above. All or part of each of the embodiments may be selectively combined to be implemented.

According to the present invention, a service provider who trades a product, a service, or a content such as an online shopping mall or an app store can recommend a product suitable for a user's inclination by appropriately reflecting a user's category preference, greatly contributing to the activation of a service, and selling a product By greatly increasing user usage, it is possible to maximize profits from services.

Claims (20)

A storage unit in which the number of times of use of each category during the reference period for each of the users is stored for each of the unit periods; And A control unit for calculating a category-specific preference corresponding to the reference period by using the recency weights for the at least one of the unit periods and the number of times of use per category Category preference apparatus comprising a. The method according to claim 1, The control unit And adding the number of times of use of each category to which the freshness weights are applied for each category to calculate a category-specific preference corresponding to the reference period in which the latestness is reflected. The method according to claim 2, The control unit Applying a first recency weight to a first unit period of the unit periods, and applying a second recency weight greater than the first recency weight to a second unit period closer to the current time point than the first unit period. Category preference apparatus characterized in that. The method according to claim 3, The freshness weight is Calculated by applying an exponential function, and adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function to adjust the degree of reduction of the freshness weights. A device for generating a category preference. The method according to claim 2, The control unit And calculating a trend line corresponding to a change in the number of times of use during the reference period, and calculating a preference for each category corresponding to the reference period in which the trend is reflected according to the slope of the trend line. The method according to claim 5, The control unit And a trend weighting factor higher than a category having a negative value is applied to a category having a positive slope. A category preference generator for calculating category preferences corresponding to the reference periods using recency weights for any one or more of the unit periods and the number of use of each category during the reference periods for each of the users; A recommendation SET generator generating the recommendation set for each category; And A product recommender that selects recommended products in the recommendation set for each category and recommends selected products by considering one or more of a tendency corresponding to a change in the number of times of use during the reference period and preferences of the categories Product recommendation apparatus comprising a. The method according to claim 7, The tendency is And a product recommendation apparatus calculated based on a slope of a trend line corresponding to a change in the number of times of use. The method according to claim 8, The product recommender And recommending a product of a category having a positive slope to the trend line over a product of a category having a negative slope to the trend line. The method according to claim 9, The product recommender A recommendation product selector for selecting recommendation products in the recommendation set for each category; A tendency generator for calculating the tendency by calculating a change in the number of times of use for each category; A recommendation rank controller for determining recommendation priorities of the recommended products based on the tendency; And An output interface for providing the recommendation products to a user according to the recommendation priority Product recommendation apparatus comprising a. The method according to claim 7, The category preference generator And calculating the preference for each category corresponding to the reference period in which the latestness is reflected by adding the number of times of use of each category to which the latest weights are applied. The method according to claim 11, The category preference generator Applying a first recency weight to a first unit period of the unit periods, and applying a second recency weight greater than the first recency weight to a second unit period closer to the current time point than the first unit period; A product recommendation device, characterized in that to calculate the preference for each category. The method according to claim 12, The freshness weight is Calculated by applying an exponential function, and adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function to adjust the degree of reduction of the freshness weights. Product recommendation apparatus characterized by the above. Storing, for each of the unit periods, the number of times of use per category during the reference period for each of the users; Calculating freshness weights for any one or more of the unit periods; And Calculating a preference for each category corresponding to the reference period by using the freshness weights and the number of times of use of each category; Category preference generation method for each user, characterized in that it comprises a. The method according to claim 14, Computing the freshness weights Computing the freshness weights such that a first freshness weight applied to a first unit period of the unit periods is less than a second freshness weight applied to a second unit period closer to the current time point than the first unit period. Category preference generation method for each user, characterized in that. The method according to claim 15, The freshness weight is Calculated by applying an exponential function, and adjusting an absolute value of a value obtained by dividing the exponential function with respect to a period divided by the exponential function to adjust the degree of reduction of the freshness weights. Characteristic preference generation method for each user characterized by. Generating a recommendation set for each category; Calculating category-specific preferences corresponding to the reference periods using recency weights for any one or more of the unit periods and the number of use of each category during the reference period for each of the users; Selecting recommended products in the recommendation set for each category according to the preference for each category; And Recommending the recommended products according to a priority corresponding to a trend corresponding to a change in the number of times of use during the reference period. Product recommendation method comprising a. The method according to claim 17, The tendency is The product recommendation method of claim 1, wherein the product is calculated based on a slope of a trend line corresponding to the change in the number of use. The method according to claim 18, The recommendation step The product recommendation method of claim 1, wherein a product of a category having a positive slope of the trend line is preferentially recommended than a product of a category having a negative value of the trend line. A recording medium having recorded thereon a program for executing the method of claim 14.

Images