JP7021475B2 - Information processing equipment and programs - Google Patents

Description

The present invention relates to an information processing apparatus and a program.

As a mechanism for recommending products and services (hereinafter referred to as items) to users, Patent Document 1 discloses that weighting is performed according to the degree of use of each item by the user in the item ranking such as the order of popularity of each item. Has been done.

Japanese Unexamined Patent Publication No. 2013-105213

Collaborative filtering, which is widely used in this type of recommender system, correlates the similarity between users or the co-occurrence between items, but tends to recommend items that are popular with a large number of users. ..

Whether or not a user tends to select a popular item when making a purchasing behavior depends on each user. In other words, some users tend to buy themselves because they are popular with other users, and some users tend to buy what they want to buy regardless of whether they are popular or not. There are various degrees of orientation.

In the past, it was not possible to recommend an item in consideration of whether or not the user tends to select a popular item in addition to the user's preference.

Therefore, an object of the present invention is to make a recommendation according to the degree of user's orientation toward popularity.

The invention according to claim 1 uses the purchase history, browsing history, or selection history of the object by the user to determine the user's preference, the characteristics of the object, and the user's orientation toward the popularity of the object. Therefore, there is a calculation unit that calculates different orientations for each group that divides the object according to a predetermined standard, the degree of relevance between the user's preference and the characteristics of the object, the popularity of the object, and the object. It is characterized by including a specific unit for specifying a recommended object, which is an object to be recommended to the user, and a presenting unit for presenting the specified recommended object to the user, based on the user's orientation toward the popularity of the. To provide an information processing device.

In the invention according to claim 2 , in the configuration according to claim 1 , the group is classified by the genre to which the object belongs.

In the invention according to claim 3 , in the configuration according to claim 1 , the group is classified according to the use of the object.

In the invention according to claim 4 , in the configuration according to claim 1 , the group is classified by the price of the object.

In the invention according to claim 5 , in the configuration according to claim 1 , the group is classified by the brand of the object.

The invention according to claim 6 uses a computer to use a user's purchase history, browsing history, or selection history of the object to the user's preference, characteristics of the object, and popularity of the object. A calculation unit that calculates the degree of orientation, which is different for each group in which the object is divided according to a predetermined standard, the degree of relevance between the user's preference and the characteristics of the object, and the degree of popularity of the object. Based on the user's orientation toward the popularity of the object, it functions as a specific unit that specifies the recommended object that is the object to be recommended to the user and a presentation unit that presents the specified recommended object to the user. Provide a program for.

According to the inventions according to claims 1 and 6 , the user's orientation toward the popularity of the object using the purchase history, browsing history, or selection history of the object by the user, which is a group classified by a predetermined standard. It is possible to make recommendations according to the degree of orientation corresponding to.
According to the invention of claim 2 , it is possible to make a recommendation according to the degree of orientation corresponding to the group classified by the genre to which the object belongs.
According to the invention of claim 3 , it is possible to make a recommendation according to the degree of orientation corresponding to the group classified according to the use of the object.
According to the invention of claim 4 , it is possible to make a recommendation according to the degree of orientation corresponding to the group classified by the price of the object.
According to the invention of claim 5 , it is possible to make a recommendation according to the degree of orientation corresponding to the group classified by the brand of the object.

The figure which shows the hardware composition of the information processing apparatus 1. The block diagram which shows the functional structure of the information processing apparatus 1. The figure explaining the mechanism of the conventional recommendation. The flowchart which shows the operation of the information processing apparatus 1. The flowchart which shows the operation of the information processing apparatus 1. The flowchart which shows the operation of the information processing apparatus 1. The flowchart which shows the operation of the information processing apparatus 1.

An example of an embodiment of the present invention will be described.
<Structure>
Hereinafter, an embodiment of the present invention will be described. The target selected by the user at the time of purchase is a product, service, or the like, but in the following, such a target that can be selected by the user is collectively referred to as an "object".

FIG. 1 is a diagram showing a hardware configuration of the information processing device 1. The information processing device 1 is an example of the information processing device according to the present invention. The information processing apparatus 1 identifies an object group recommended to the user, and presents information about the object group to the user. The information processing device 1 is a computer including a control unit 11, a storage unit 12, a communication unit 13, and a UI (User Interface) unit 14.

The control unit 11 includes an arithmetic unit such as a CPU (Central Processing Unit) and a storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory), and controls the entire information processing device 1. conduct. The ROM stores the firmware that describes the procedure for booting the hardware and the OS (Operating System). The RAM is used to store data when the CPU executes an operation.

The storage unit 12 is provided with, for example, a semiconductor memory or a hard disk storage device, and stores an OS (Operation System) and programs for realizing various functions. Further, the storage unit 12 stores information (purchase history) regarding the result of a plurality of users using a shopping website on the Internet by using a communication device such as a personal computer or a smart phone. The purchase history includes, for example, the date and time when the object was purchased, the user ID which is the identification information of the user who purchased the object, and the object ID which is the identification information of the purchased object. These purchase histories may be collected in the information processing device 1 via the Internet from the web server device that manages the shopping website on the Internet. Further, the information processing device 1 itself may be a web server device that manages a shopping website on the Internet.

The communication unit 13 is provided with a communication I / F (Interface) for communicating with an external communication device, and is connected to a communication line such as the Internet. The UI unit 14 includes a display unit and an operation unit. The display unit includes, for example, a liquid crystal display device, and displays a screen for operating the information processing device 1 on the display surface. The operation unit is provided with, for example, a touch panel provided so as to cover the display surface of the display unit, a keyboard provided at a position adjacent to the display surface, and the like, and receives an operation and controls a signal according to the operation. Output to unit 11. The control unit 11 controls the information processing device 1 according to the content of this operation.

Here, an example of an algorithm for recommending an object that matches the user's taste to the user will be briefly described. In this algorithm, the user is based on the premise that an object similar to an object purchased by the user is preferred by the user, and an object similar to an object not purchased by the user is not preferred by the user. Learn about your tastes.
The strength vui of the user u's preference for the object i is expressed by the equation (1).

Here, qi T (T is a transposed matrix) is a vector showing the characteristics of the object (object feature vector), and pu is the user's preference, that is, the vector showing the characteristics of the object preferred by the user (user preference vector). Is. The inner product of these vectors is the degree of matching between the preference of the user u and the feature of the object i, and corresponds to the strength vui of the preference of the user u with respect to the object i.

The difference Xuij between the user u's preference for the object i and the user u's preference for the object j Xuij is expressed by the equation (2).

The probability p that the user u purchases the object i and does not purchase the object j is expressed by the above-mentioned equation (3) which is a sigmoid function of xuij. Here, I is a complete set of objects, and Iu + is a subset of the objects purchased by the user u among the total sets of objects.

Then, the parameter Θ is sequentially updated and learning is performed so that the probability is maximized as shown in the equation (4). The parameter Θ includes, for example, each component of each user's vector pu and each item's qi. That is, learning is performed so that the difference between the strength of the preference for the object purchased by the user and the strength of the preference for the object not purchased by the user becomes large.

Here, the popularity of a product fluctuates with the passage of time. Further, when a user makes a purchasing action, whether or not the user tends to select a popular item at the timing of the action depends on the user. In other words, some users tend to buy themselves because they are popular with other users, and some users tend to buy what they want to buy regardless of whether they are popular or not. Is various. In this way, the index value indicating whether or not each user tends to select a popular object is the degree to which the user is oriented toward the popularity of the object. ".

When recommending an object to a user, it is desirable to consider the user's popularity orientation as well as the user's preference. Therefore, in the present embodiment, the user can identify the object recommended to the user by analyzing the user's preference and popularity orientation based on the time-based popularity of the object and the user's behavior history. Recommend matching objects.

FIG. 2 is a block diagram showing a functional configuration of the information processing apparatus 1. These functions are realized by the control unit 11 executing the program stored in the storage unit 12. The acquisition unit 101 acquires the purchase history for each user. Using the purchase history, the calculation unit 102 uses a user preference vector that expresses each user's preference in a vector format, an object feature vector that expresses the characteristics of each object in a vector format, and a popularity degree of each object. The popularity of the object and the popularity of the object, which is the orientation of each user with respect to the popularity of the object, are calculated. In addition to the above-mentioned purchase history, the storage unit 103 stores these user preference vectors, object feature vectors, object popularity, and popularity orientation. The specifying unit 104 identifies a recommended object, which is an object recommended to the user, based on the relevance of the user preference vector and the object feature vector, the object popularity, and the popularity orientation. Here, the specifying unit 104 acquires the popularity of the object, which changes with the passage of time, at the timing of specifying the recommended object. That is, the specific unit 104 acquires the latest popularity of the object. The presentation unit 105 presents the specified recommended object to the user.

<Operation>
Next, the operation of the information processing apparatus 1 will be described, but before that, an example of the conventional recommendation method will be described with reference to FIG.

In FIG. 3, elements 1 to 8 are components of a user preference vector and an object feature vector. For example, when the object is a movie, the elements 1 to 8 are a comedy element, a horror element, an educational element, a historical element, an SF (Science Fiction) element, a romantic element, and the like. .. FIG. 3 exemplifies the value of each element for a certain user U1's preference (upper row) and the value of each element for a certain object A1 (lower row), and the larger the value, the more the corresponding vector. It means that the component is large, and the smaller the value, the smaller the corresponding vector component.

The degree to which the preference of the user U1 matches the feature of the object A1 is expressed by the inner product of the user preference vector (upper part of FIG. 3) and the object feature vector (lower part of FIG. 3). That is, it is a value obtained by multiplying the component of the user preference vector corresponding to the same element and the component of the object feature vector, and adding them together. In the example of FIG. 3, 0.1 × 0.5 + (−0.2) × (−0.2) + 0.3 × 0.7 + (−0.4) × 0.1 + (−0.1) × (-0.1) + 0.5 × 0.1 + (-0.2) × 0.8 + 0.3 × 0.1 = 0.19 is the degree to which the preference of the user U1 matches the characteristics of the object A1. Become.

Further, when the popularity of the object A1 calculated separately is, for example, 0.15, the degree of recommending the object A1 to the user U1 is 0.15 (object popularity) + 0.19 (user U1). The degree to which the preference of A1 matches the characteristics of the object A1) = 0.34. The relationship between the popularity of the object and the degree to which the preference of the user U1 matches the characteristics of the object A1 is not limited to the simple addition relationship as described above, but is a relationship using, for example, multiplication or division. You may. However, in any case, such a conventional method tends to recommend an object that is popular with a large number of users.

From FIG. 4, the processing procedure of the information processing apparatus 1 in the present embodiment is shown. In FIG. 4, the calculation unit 102 calculates the user preference vector, the object feature vector, the object popularity, and the popularity orientation using the purchase history stored in the storage unit 103 (step A). S100).

Here, FIG. 5 is a diagram illustrating the details of the process A. First, the calculation unit 102 initializes the user preference vector Pu, the object feature vector qi, the popularity orientation p, and the object popularity b, which are the above-mentioned parameters, and determines the number of learning times E (step S101). Next, the calculation unit 102 acquires the history number N of the purchase history stored in the storage unit 103 (step S102), and substitutes 1 for the variable e (step S103). When the variable e is smaller than the number of learning times E (step S104; YES), the calculation unit 102 calculates the user preference vector Pu, the object feature vector qi, the popularity orientation p, and the object popularity b (process). A1, step S105).

Here, FIG. 6 is a diagram illustrating the details of the process A1. First, the calculation unit 102 substitutes 1 for the variable n (step S1051). When the variable n is smaller than the history number N (step S1052; YES), the calculation unit 102 has the user preference vector Pu, the object feature vector qi, which are parameters corresponding to the user u, the object i, and the time t. With reference to the popularity-oriented degree p and the object popularity degree b (step S1053), Xuij is calculated according to the above-mentioned equations (1) to (4), and the log-likelihood for each parameter is calculated (steps S1054 and S1055). ). Then, the calculation unit 102 uses this log-likelihood to update each parameter, the user preference vector Pu, the object feature vector qi, the popularity orientation p, and the object popularity b (step S1056).

Here, as a method of calculating the popularity of the object, for example, as shown in the formula (5), the total number of sales (number of purchases) of a certain object (product in the formula) i at a certain time t is totaled. , An example of calculating the popularity by multiplying the aggregated value by the weighting coefficient wt at time t can be considered. The function for calculating the popularity of the object may be any monotonic decreasing function with the number of user actions (for example, purchasing a product or selecting a product) as a variable.

Further, as a method of calculating the popularity orientation, for example, as shown in the formula (6), an example of calculating the ratio of purchasing an object having a popularity higher than the threshold value in the past purchase history of the user u can be considered. Be done. However, the function for calculating the popularity orientation is not limited to this example.

In this way, the process of FIG. 6 is repeated N times (steps S1057 and S1052).

Returning to the description of FIG. 5, the above process is repeated E times (steps S106 and S104).

Returning to FIG. 4, the specifying unit 104 identifies the recommended object, which is the object recommended to the user, based on the relevance of the user preference vector and the object feature vector, the object popularity, and the popularity orientation. (Process B, step S110).

Here, FIG. 7 is a diagram illustrating the details of the process B. First, the specific unit 104 acquires all the objects M (step S1101) and substitutes 1 for the variable m (step S1102). When the variable m is smaller than the number of products M (step S1103; YES), the specific unit 104 refers to the user preference vector Pu, the object feature vector qi, the popularity orientation p, and the object popularity b (step S1104). ), A recommendation value indicating the strength recommended to the user is calculated (step S1105).

For example, in the example of FIG. 3, when the user's orientation to the popularity is 0.25, the object popularity is multiplied by the popularity orientation to 0.15 (object popularity) × 0.25 (object popularity). A recommendation value of 0.38 (the degree to which the preference of the user U1 matches the characteristics of the object A1) = 0.38 (hereinafter referred to as the former example) is calculated. When the user's orientation to popularity is -0.15, 0.15 (object popularity) x (-0.15) (user's popularity orientation) + 0.19 (user U1 preference). Calculates the recommended value of (degree of matching with the characteristics of the object A1) = 0.31 (hereinafter, referred to as the latter example).

Compared to the latter example, the former example is an example of a user who has a strong tendency to buy himself because it is popular with other users, and the recommended value is relatively large. That is, the specific unit 104 determines the degree of relevance of the user's preference and the feature of the object to the user whose popularity orientation degree is higher than a predetermined value (“0” in the above example). In comparison, the recommendation target is specified with an emphasis on the degree of preference for popularity.

On the other hand, the latter example is an example of a user who is less popular than the former example but has a strong tendency to buy one that suits his / her taste, and the recommended value is relatively small. That is, the specific unit 104 refers to a user having a high degree of popularity orientation of the user less than a predetermined value (“0” in the above example) as compared with the degree of orientation toward the popularity of the user. The recommended object is specified with an emphasis on the degree of relevance between the user's preference and the characteristics of the object.

In this way, the process of FIG. 7 is repeated M times (steps S1106 and S1103). Then, the specific unit 104 extracts L objects that the user has not purchased in the past in descending order of the recommended value, and creates a recommendation list including information on these (for example, the name and price of the object). (Step S1107).

Returning to FIG. 4, the presentation unit 105 presents the created recommendation list to the user by transmitting it to the user's communication device and displaying it (step S120).

According to this embodiment, it is possible to make a recommendation according to the degree of preference of the user with respect to the degree of popularity.

<Modification example>
The embodiment may be modified as follows. Moreover, you may combine a plurality of modification examples.
<1>
The degree of orientation toward the popularity of the user may be different for each group in which the object is classified according to a predetermined standard. For example, even for the same user, the degree of popularity for movies, the degree of popularity for food and drink, and the degree of popularity for clothing are different. The groups at this time can be classified from various viewpoints. For example, the group may be divided by the genre to which the object belongs. The genre is, for example, a movie, food and drink, clothing, a book, and the like, and can be said to be a category to which the object belongs. Further, for example, the group may be classified according to the use of the object (for example, daily goods and luxury goods, practical goods and hobby goods, etc.). Further, for example, the group may be classified by the price of the object (for example, high price range, medium price range, low price range, etc.). Further, for example, the group may be classified by the brand of the object (for example, a luxury brand and other brands, or a famous brand and other brands). Regarding the degree of popularity orientation for each of these groups, the calculation unit 102 may execute the above-mentioned processes of FIGS. 4 to 7 for each group to which the object belongs.

<2>
As the user's action history, not only the purchase history described in the embodiment but also the browsing history of the object and the selection history such as putting the object in the shopping cart or favorite on the web page may be used. That is, the calculation unit 102 can calculate the user's preference for the user's preference, the feature of the object, or the popularity of the object by using the purchase history, browsing history, or selection history of the object by the user.
Further, in the embodiment, the user has learned the preference of the user based on the result of using the shopping website on the Internet, but based on the result of using the actual store selling the actual object. You may learn the preference of the user.

<3>
The object popularity may be an index value of the object popularity, and the calculation method thereof is not limited to the example of the embodiment. Further, the popularity orientation may be an index value of the degree of preference for the popularity of the object, and the calculation method thereof is not limited to the example of the embodiment.

<4>
In the above embodiment, an example in which the control unit 11 realizes the above function by executing a program is shown, but the above function may be implemented in a hardware circuit. Further, the program may be recorded and provided on a computer-readable recording medium such as an optical recording medium or a semiconductor memory, and the program may be read from the recording medium and installed in the information processing apparatus. In addition, this program may be provided by a telecommunication line.

1 ... Information processing device, 11 ... Control unit, 12 ... Storage unit, 13 ... Communication unit, 14 ... UI unit, 101 ... Acquisition unit, 102 ... Calculation unit, 103 ... Storage unit, 104 ... Specific unit, 105 ... Presentation unit ..

Claims (6)

Using the purchase history, browsing history, or selection history of the object by the user, the user's preference, the characteristics of the object, and the user's orientation toward the popularity of the object are defined. A calculation unit that calculates different orientations for each group classified according to the criteria of
Identify the recommended object to be recommended to the user based on the relevance of the user's preference and the characteristics of the object, the popularity of the object, and the user's orientation toward the popularity of the object. With a specific part to do
An information processing apparatus including a presentation unit that presents the specified recommended object to the user. The information processing apparatus according to claim 1 , wherein the group is classified by a genre to which the object belongs. The information processing apparatus according to claim 1 , wherein the group is classified according to the use of the object. The information processing apparatus according to claim 1 , wherein the group is classified by the price of the object. The information processing apparatus according to claim 1 , wherein the group is classified by the brand of the object. Computer,
Using the purchase history, browsing history, or selection history of the object by the user, the user's preference, the characteristics of the object, and the user's orientation toward the popularity of the object are defined. A calculation unit that calculates different orientations for each group classified according to the criteria of
Identify the recommended object to be recommended to the user based on the relevance of the user's preference and the characteristics of the object, the popularity of the object, and the user's orientation toward the popularity of the object. With a specific part to do
A program for functioning as a presenting unit that presents the specified recommended object to the user.

Images