JP7423994B2 - Recommendation device and recommendation method - Google Patents

Description

The present invention relates to a recommendation device and a recommendation method.

EC (electronic commerce) sites using the Internet etc. propose products that are expected to be of interest to the user, based on information about the products the user has previously purchased or viewed. For example, see Patent Document 1).

Japanese Patent Application Publication No. 2014-160396

As mentioned above, in the past, products related to the pages and products that the user viewed on the Internet were proposed. Therefore, the range of proposed products is limited to those related to the purchased or viewed products. Furthermore, the products purchased by the user or the products whose information is viewed may not be for the user. In this case, the above-mentioned conventional technology may not be able to suggest products that match the customer's tastes.

The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a recommendation device and a recommendation method that can recommend products that match the customer's tastes from among a wide range of products.

One aspect of the present invention that solves the above-mentioned problems includes: an information acquisition unit that acquires analysis information that is one or more information regarding at least one of preferences and intentions; and the analysis information acquired by the information acquisition unit. a calculation unit that calculates a preference type score that quantitatively represents the degree to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the calculation unit; and the preference type score calculated by the calculation unit. and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds, a selection unit that selects the product to be recommended to the customer and outputs information regarding the selected product; Based on the preference type score of the customer calculated by the calculation unit and the preference category information of the product selected by the customer, the selection unit makes a recommendation based on the customer's preference type score and the preference category information. A first learning unit that changes the value of a parameter used in an algorithm for selecting the product to be selected .
Further, one aspect of the present invention for solving the above-mentioned problems includes: an information acquisition unit that acquires analysis information that is one or more information regarding at least one of customer preferences and intentions; a calculation unit that calculates a preference type score that quantitatively represents the extent to which each of a plurality of preference categories used as elements representing preference trends matches the customer based on the analysis information; Selecting the product to be recommended to the customer based on the preference type score and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds, and outputting information regarding the selected product. It is characterized by comprising a selection section, and a second learning section that changes the preference category information based on the preference type score of the customer calculated by the calculation section and the product selected by the customer. This is a recommendation device.

One aspect of the present invention is the above-mentioned recommendation device, wherein the information acquisition unit further acquires analysis information that is information related to the customer's attributes or biometric information.

One aspect of the present invention is the recommendation device described above, wherein the product information includes information on the classification of the product, and the selection unit recommends the product to the customer from among the products in the category selected by the customer. The method is characterized in that the product is selected.

One aspect of the present invention is the recommendation device described above, wherein the selection unit selects the product to be recommended to the customer from among the products in a classification different from the classification selected by the customer. shall be.

One aspect of the present invention is the above-mentioned recommendation device, wherein the preference category information is represented by a value that quantitatively represents the extent to which each of the plurality of preference categories matches the product. .

One aspect of the present invention is the above-mentioned recommendation device, wherein the information acquisition unit presents a question to the customer to obtain, as an answer, the analysis information regarding at least one of the customer's preferences and intentions. , is characterized by.

One aspect of the present invention is a recommendation method executed by a recommendation device, which includes an information acquisition step of acquiring analysis information that is one or more information regarding at least one of customer preferences and intentions; a calculation step of calculating a preference type score that quantitatively represents the extent to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the analysis information acquired in the step; a selection step of selecting the product to be recommended to the customer based on the preference type score calculated in step and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds; an output step of outputting information regarding the product selected in the selection step, the preference type score of the customer calculated in the calculation step, and the preference category information of the product selected by the customer; In the selection step, the first learning unit changes values of parameters used in an algorithm for selecting the product to be recommended based on the customer's preference type score and the preference category information. Features.

One aspect of the present invention is a recommendation method executed by a recommendation device, which includes an information acquisition step of acquiring analysis information that is one or more pieces of information regarding either customer preferences or intentions; a calculation step of calculating a preference type score that quantitatively represents the extent to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the acquired analysis information; a selection step of selecting the product to be recommended to the customer based on the calculated preference type score and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds; an output step of outputting information regarding the product selected in the calculation step, and a second learning based on the preference type score of the customer calculated in the calculation step and the product selected by the customer. It is characterized in that the section changes the preference category information .

According to the present invention, it is possible to recommend products that match the customer's preferences from a wide range of products.

1 is a diagram showing the overall configuration of a recommendation system in this embodiment. It is a block diagram of a recommendation device in this embodiment. It is a figure showing an example of customer information in this embodiment. It is a figure showing an example of score information by answer in this embodiment. It is a figure showing an example of score information by attribute in this embodiment. It is a figure showing an example of score information classified by biometric information in this embodiment. It is a figure showing an example of product information in this embodiment. It is a flow diagram showing processing of a recommendation device in this embodiment. It is a figure which shows the example of a display of the question display screen in this embodiment. It is a figure which shows the example of a display of the preference type display screen in this embodiment. FIG. 2 is a diagram illustrating an example of a business format to which the recommendation system according to the present embodiment can be applied.

Embodiments of the present invention will be described in detail below with reference to the drawings.
FIG. 1 is a diagram showing the overall configuration of a recommendation system 1. As shown in FIG. The recommendation system 1 includes a recommendation device 2, a customer information storage device 3, a preference judgment material storage device 4, and a product information provision device 5 of an information provider, a client terminal 6 and a web server 7, and a customer terminal 8 of a customer. Equipped with. The customer information storage device 3, the preference judgment material storage device 4, and the product information providing device 5 are connected to the recommendation device 2. Any two or more of the recommendation device 2, customer information storage device 3, preference judgment material storage device 4, and product information providing device 5 may be integrated into one device. The recommendation device 2, client terminal 6, web server 7, and customer terminal 8 are connected to a network 9 such as the Internet. One or more of the customer information storage device 3, the preference judgment material storage device 4, and the product information providing device 5 may be connected to the network 9. The client is a company that sells products. The product may be a physical product such as a product, or may be a variety of services. Customers use clients to purchase products or consider purchasing. The number of clients and the number of customers are arbitrary. In the figure, one customer information storage device 3, one preference judgment material storage device 4, one product information providing device 5, one client terminal 6, one web server 7, and one customer terminal 8 are shown, but the number of each is arbitrary. . Further, the client may have the customer information storage device 3 and the product information providing device 5.

The recommendation device 2 acquires one or more pieces of analysis information from a client terminal 6 or a customer terminal 8 connected to a web server 7 . The analysis information is information regarding at least one of the customer's preferences and orientations. The recommendation device 2 may further acquire information regarding customer attributes or biometric information as analysis information. Information on customer preferences, orientations and attributes is input by the customer or client into a client terminal 6 or by the customer into a customer terminal 8 connected to a web server 7. The biometric information is collected by a biometric information acquisition device included in the client terminal 6 or the customer terminal 8. As the biometric information acquisition device, in addition to a dedicated biometric information measuring device, a camera or the like can also be used. The recommendation device 2 may read out part of analysis information such as customer attributes from the customer information storage device 3. The biological information includes, for example, facial expression, skin color, brain waves, pulse, breathing, and sweating. Biometric information is used in addition to or in place of information regarding customer preferences or preferences.

The recommendation device 2 determines preference types that classify trends in customer preferences based on the acquired analysis information. The preference type is expressed based on a preference type score consisting of a numerical value that quantitatively represents the extent to which each of a plurality of different preference categories matches the customer. In other words, the preference category is used as an element representing trends in customer preferences. The recommendation device 2 selects products to recommend to the customer based on the preference category associated with each product and the customer's preference type score. The recommendation device 2 outputs information on the selected product to the client terminal 6 or to the customer terminal 8 connected to the web server 7, and displays the information.

The customer information storage device 3 stores customer information indicating information regarding customers. The customer information may include information on customer attributes. The preference judgment material storage device 4 stores preference judgment materials, score calculation information, and type classification information. The preference determination material is data on materials presented to obtain information on the customer's preferences and intentions. The score calculation information indicates a method of calculating a preference type score according to the analysis information. The product information providing device 5 stores product information of each product sold by the client. The product information includes information on preference categories corresponding to the product.

The client terminal 6 is, for example, a computer terminal such as a PC (personal computer) installed at a client's physical store or a digital signage. The web server 7 is a computer server that provides web pages for clients. The web server 7 provides, for example, a web page of a client's corporate site or an EC site that sells the client's products. The customer terminal 8 is a computer terminal such as a PC, a smartphone, or a tablet terminal used by the customer. The customer terminal 8 displays a web page provided by the web server 7.

FIG. 2 is a block diagram showing the internal configuration of the recommendation device 2. As shown in FIG. The recommendation device 2 includes a communication section 210, a storage section 220, and a processing section 230.

The communication unit 210 communicates with other devices via the network 9. The communication unit 210 includes a receiving unit 211 that receives information from other devices, and a transmitting unit 212 that transmits information to the other devices. The storage unit 220 stores various types of information used by the processing unit 230. The storage unit 220 stores information read from the customer information storage device 3, the preference judgment material storage device 4, and the product information providing device 5, as well as information such as the preference type score and preference type of each customer.

The processing unit 230 includes a recommendation unit 240 and a learning unit 250. The recommendation section 240 includes an information acquisition section 241 , a calculation section 242 , an analysis result output section 243 , and a selection section 244 . The information acquisition unit 241 acquires analysis information from the customer information storage device 3 , the client terminal 6 , or the customer terminal 8 connected to the web server 7 . The analysis information is information regarding at least one of the customer's preferences, intentions, attributes, and biometric information. The calculation unit 242 uses the analysis information acquired by the information acquisition unit 241 and the score calculation information stored in the preference judgment material storage device 4 to quantify the extent to which the customer matches each of a plurality of different preference categories. Calculate the preference type score expressed as follows. The analysis result output unit 243 outputs information on the preference type score to the client terminal 6 or the customer terminal 8 connected to the web server 7. The selection unit 244 selects products to recommend to the customer based on the preference type score calculated by the calculation unit 242 and the preference category to which each product corresponds, and sends information on the selected products to the client terminal 6 or to the web server. It is output to the customer terminal 8 connected to 7. The selection unit 244 reads information on the preference category to which each product corresponds from the product information providing device 5.

The learning section 250 includes a first learning section 251 and a second learning section 252. The first learning unit 251 uses the information of the customer's preference type score calculated by the calculation unit 242 and the preference category corresponding to the product selected by the customer as a purchase target as learning data, and adjusts the parameters used in the selection algorithm. Learn values. The selection algorithm is an algorithm used by the selection unit 244 to select products to recommend based on the preference type score and the preference category to which each product corresponds. The second learning unit 252 uses the customer's preference type score calculated by the calculation unit 242 and the product selected by the customer as a purchase target as learning data to learn a preference category corresponding to the product.

The recommendation device 2 is realized by, for example, one or more computer devices. When the recommendation device 2 is realized by a plurality of computer devices connected through a network, which functional unit is realized by which computer device can be arbitrarily determined. For example, the recommendation unit 240 and the learning unit 250 may be implemented by different computer devices. Further, one functional unit may be realized by a plurality of computer devices.

FIG. 3 is a diagram showing an example of customer information stored in the customer information storage device 3. As shown in FIG. The customer information includes information such as each customer's customer ID, name, address, date of birth, gender, occupation, family, purchase history, and preference type score. The occupation information may include position information. Further, the family information includes information on whether the person is married or single, and information on the family structure. From the customer information, attribute information representing the customer's attributes, which is used as analysis information, is obtained. For example, attribute information about a residential area can be obtained from an address. Further, attribute information such as the customer's age and generation can be obtained from the date of birth. Furthermore, information on gender, occupation, and family is used as attribute information as is. The purchase history indicates products that the customer has purchased in the past. Furthermore, the customer information may include attribute information such as annual income and disposable income, which is used as analysis information.

The preference judgment material storage device 4 stores score information by answer, score information by attribute, and score information by biometric information as information for score calculation.

FIG. 4 is a diagram showing an example of score information by answer. The score information by answer indicates the score by preference category given to the customer according to the customer's answer to the question for acquiring analysis information regarding the customer's preferences and orientation. The answer-specific score information is information that associates the question number, the template used for the question display screen, the preference judgment material, and the calculation information for each answer.

The template is used to generate screen data for a question display screen that outputs preference judgment materials and displays questions such as whether you like or dislike the output preference judgment materials, or whether you like any of the plurality of output preference judgment materials. This is the data used. The preference determination materials include image data, video data, text data, audio data, and the like. As the preference judgment material, for example, it is possible to use something that makes it easy to intuitively select whether you like it or not (YES/NO), or something that makes it easy to intuitively select which one you like. The information on the preference judgment material included in the score calculation information may be the storage location of the preference judgment material in the preference judgment material storage device 4 or another device.

The calculation information indicates the calculation to be performed on the customer's preference type score in response to the answer to the question. In addition to the two-choice answer, the answer may be one in which the degree of liking or disliking is selected from three or more levels, or one may be selected from three or more options. The calculation information indicates the value of each preference category to be added, subtracted, or multiplied. When only addition is performed, information on the type of operation can be omitted. Only one preference category may be the object of calculation for one answer, or a plurality of preference categories may be the object of calculation for one answer. The calculation information may indicate a question number and a value to be multiplied by the value of each preference category to be added according to the answer to the question with that question number. The value corresponding to each preference category of the calculation information corresponds to the value of the parameter used in the selection algorithm.

FIG. 5 is a diagram showing an example of score information by attribute. The attribute-specific score information is information that associates attribute values of attribute information with calculation information. The calculation information is expressed in the same way as the calculation information included in the score information by answer.

FIG. 6 is a diagram showing an example of score information classified by biometric information stored in the preference judgment material storage device 4. The score information by biometric information indicates a score by preference category that is given to a customer according to changes in the customer's biological body in response to a question for acquiring analysis information regarding the customer's preferences and orientation. Biometric information can be used instead of the customer inputting an answer to the question of whether they like or dislike the preference judgment material output on the question display screen, or whether they like any of the plurality of preference judgment materials output, or It can be used as additional information for the input answer. For example, it is possible to determine a customer's preferences from biological information such as a change in the customer's facial expression when viewing an image of preference determination material. The biometric information-specific score information is information that associates the question number, the template used for the question display screen, the preference determination material, and the calculation information by biometric information. The question number, template, preference determination material, and calculation information are the same as the question number, template, preference determination material, and calculation information included in the score information by answer. Biometric information is information that associates the type of biometric information with the value of that type. If multiple types of biometric information are associated with one question number, calculation information corresponding to any type of biometric information may be used to calculate the preference type score, and multiple types of biometric information may be used to calculate the preference type score. Calculation information corresponding to each may be used to calculate the preference type score.

FIG. 7 is a diagram showing an example of product information stored in the product information providing device 5. As shown in FIG. The product information includes the product ID, product classification, product name, product image, preference category, product grade, product description, sales information, etc. of each product. Product classification may be hierarchical. For example, product classifications can be hierarchical, such as "clothing - women's - outerwear", "household - miscellaneous goods - rugs", etc. The occupation information may include position information. Further, the family information includes information on whether the person is married or single, and information on the family structure. The sales information includes information on dealers indicating clients who handle the product, and the price and delivery date of the product at each dealer.

FIG. 8 is a flow diagram showing the processing of the recommendation device 2. In the following, a case where the recommendation device 2 receives access from the client terminal 6 will be described as an example.

First, the information acquisition unit 241 of the recommendation device 2 receives information identifying the client from the client terminal 6, and the customer ID and attribute information input by the operator. The operator is the customer or client. The information acquisition unit 241 reads customer information specified by the customer ID from the customer information storage device 3, and acquires attribute information used as analysis information from the read customer information (step S105). Note that the information acquisition unit 241 may receive all attribute information from the client terminal 6, or may acquire all attribute information from customer information. When receiving all attribute information from the client terminal 6, the information acquisition unit 241 does not need to receive the customer ID. Further, when customer information is not stored in the customer information storage device 3, the information acquisition unit 241 receives information regarding the customer input by the operator from the client terminal 6, and generates customer information using the received information. It may also be written into the customer information storage device 3.

Further, the information acquisition unit 241 receives product classification information input by the operator from the client terminal 6 (step S110). Note that the information acquisition unit 241 may receive the product classification information before step S105, or may receive it together with the information received in step S105.

The information acquisition unit 241 initializes the score of each preference category making up the preference type score to 0 (step S115). The information acquisition unit 241 initializes the number of questions n by setting it to 1 (step S120).

The information acquisition unit 241 selects the score information by answer for the question number that has not been selected yet from the score information by answer stored in the preference judgment material storage device 4. The information acquisition unit 241 reads templates and preference determination materials from the answer-by-answer score information of the selected question number. The information acquisition unit 241 generates screen data of a question display screen in which preference determination materials are embedded in a template, and transmits it to the client terminal 6 (step S125). The client terminal 6 displays a question display screen based on the received screen data. Preference determination materials are presented to the customer on the question display screen.

The client terminal 6 transmits information on the answer input by the operator on the question display screen. The information acquisition unit 241 of the recommendation device 2 receives response information from the client terminal 6 (step S130). The calculation unit 242 reads calculation information according to the received answer from the answer-by-answer score information of the selected question number, and performs the calculation indicated by the read calculation information on the preference type score (step S135).

The information acquisition unit 241 determines whether the number of questions n has reached the total number of questions N (step S140). When determining that the total number of questions N has not been reached (step S140: NO), the information acquisition unit 241 adds 1 to the number of questions n (step S145). The information acquisition unit 241 repeats the processing from step S125.

If the information acquisition unit 241 determines that the number of questions n has reached the total number of questions N (step S140: YES), the calculation unit 242 performs the process of step S150. The calculation unit 242 reads calculation information corresponding to each attribute information acquired in step S105 from the attribute-specific score information stored in the preference judgment material storage device 4. The calculation unit 242 performs the calculation indicated by the read calculation information on the preference type score (step S150). The calculation unit 242 sets the calculated preference type score in the customer information stored in the customer information storage device 3.

The analysis result output unit 243 generates screen data of a preference type display screen that displays the customer's preference type score, and transmits it to the client terminal 6 (step S155). The client terminal 6 displays a preference type display screen based on the received screen data.

The selection unit 244 selects a product to recommend to the customer (step S160). Specifically, the selection unit 244 reads out a predetermined number of preference categories with high scores based on the customer's preference type scores. The selection unit 244 specifies, from among the product information stored in the product information providing device 5, product information of a product that belongs to the product classification acquired in step S110 and is associated with the read preference category. Furthermore, the selection unit 244 reads out, from the product information providing device 5, product information in which the client indicated by the identification information received from the client terminal 6 is included in the store, from among the specified product information, as recommended product information. The selection unit 244 may further read product information whose stores include other clients affiliated with the client as recommended product information. Note that the selection unit 244 may further set as a condition for reading out the recommended product information from the recommendation device 2 that a product grade corresponding to the annual income, disposable income, and position read from the customer information is set in the product information. Note that the selection unit 244 may receive product grade information input by the operator from the client terminal 6.

The selection unit 244 generates screen data for a recommendation screen that displays recommended product information, for example, in the order of recommendations. The selection unit 244 gives a higher recommendation ranking to a product that corresponds to a preference category with a higher score. The recommendation screen displays at least part of recommended product information. For example, information such as product name, product image, preference category, product grade, product description, price, and delivery date is displayed on the recommendation screen. The selection unit 244 acquires price and delivery date information from sales information of a dealer by the client using the client terminal 6 (or another client affiliated with the client). The selection unit 244 transmits the generated screen data to the client terminal 6 (step S165). The client terminal 6 displays a recommendation screen based on the received screen data.

The product information may further include information on recommended attributes corresponding to each product. Recommendation target attributes indicate attribute information of the customer to be recommended, such as age, gender, and area of residence. In this case, the selection unit 244 may further set the condition that the recommendation target attribute includes customer attribute information as the recommended product information from the recommendation device 2. Further, in step S125, the information acquisition unit 241 may select score information by answer in order of question number.

Note that instead of the user inputting an answer to the question displayed on the question display screen, the user's biometric information may be collected. In this case, the recommendation device 2 operates as follows in steps S125 to S135. That is, the information acquisition unit 241 selects the biometric information-specific score information of the question number that has not been selected yet from the biometric information-specific score information stored in the preference judgment material storage device 4. The information acquisition unit 241 reads templates and preference determination materials from the biometric information-specific score information of the selected question number. The information acquisition unit 241 generates screen data of a question display screen in which preference determination materials are embedded in a template, and transmits it to the client terminal 6 (step S125). The client terminal 6 displays a question display screen based on the received screen data. Preference determination materials are presented to the customer on the question display screen. The client terminal 6 transmits the customer's biometric information measured while displaying the question display screen to the recommendation device 2, and the information acquisition unit 241 of the recommendation device 2 receives the biometric information from the client terminal 6 (step S130). . The calculation unit 242 reads the calculation information according to the received biological information from the biological information-specific score information of the selected question number, and performs the calculation indicated by the read calculation information on the preference type score (step S135).

Furthermore, in addition to obtaining answers input by the user to the display on the question display screen, the recommendation device 2 may also collect the user's biometric information. In this case, the recommendation device 2 operates as follows in steps S130 to S135. That is, the client terminal 6 transmits information on the answer input by the operator to the question display screen and the customer's biometric information measured while the question display screen is displayed. The information acquisition unit 241 of the recommendation device 2 receives response information and biometric information from the client terminal 6 (step S130). The calculation unit 242 reads calculation information corresponding to the received answer from the answer-by-answer score information of the selected question number, and performs the calculation indicated by the read calculation information on the preference type score. Furthermore, the calculation unit 242 reads calculation information corresponding to the received biometric information from the biometric information-specific score information of the selected question number, and performs the calculation indicated by the read calculation information on the preference type score (step S135). .

In step S160, the selection unit 244 selects, among the product information stored in the product information providing device 5, products that belong to a product classification different from the product classification obtained in step S110 and are associated with the read preference category. Product information may also be specified. Then, the selection unit 244 reads out, from the product information providing device 5, product information whose store includes the client indicated by the identification information received from the client terminal 6, from among the specified product information, as recommended product information. This makes it possible to recommend products even in product categories that the customer is not aware of. Different product classifications may be determined randomly, for each client, or may be determined based on information on product classifications that are associated with each other in advance, for example. This makes it possible to recommend appliances, food, etc. to customers looking for furniture. Note that the recommended product may be a specific product or may be a product group including a plurality of related products (for example, a specific brand or genre).

In the above description, an example has been described in which there is one preference category corresponding to a product, but a product may be associated with a plurality of preference categories. In this case, a preference type score is set in the product information instead of a preference category. In step S160, the selection unit 244 specifies product information in which the product classification acquired in step S110 is set and the client indicated by the identification information received from the client terminal 6 is included in the store. The selection unit 244 calculates the inner product or angle between the vector represented by the taste type score set in the specified product information (an N-dimensional vector whose dimension number is the number of taste categories N) and the vector represented by the customer's taste type score. It may be determined that the higher the calculated inner product or the smaller the calculated angle, the higher the recommendation rank. The selection unit 244 reads out a predetermined number of product information from the product information providing device 5 as recommended product information, starting from the highest ranking product information.

Although the case where the client terminal 6 is used has been described above, the client terminal 6 may be replaced with a customer terminal 8 accessing a website provided by the web server 7. In this case, the recommendation device 2 transmits and receives data to and from the customer terminal 8 via the web server 7.

The client terminal 6 may transmit the customer ID and information on the product purchased by the customer to the recommendation device 2. For example, the client terminal 6 may cooperate with a store's POS (Point Of Sale) system to obtain information on products purchased by the customer, and the client may input information on products purchased by the customer into the client terminal 6. It's okay. The processing unit 230 of the recommendation device 2 writes the product ID of the received product into the purchase history of the customer information specified by the customer ID. The selection unit 244 reads the preference category corresponding to the product purchased by the customer from the product information, uses the read preference category as a preference category with a high score read based on the customer's preference type score, and selects the preference category in step S160. Then, the process of step S165 may be performed. Alternatively, the selection unit 244 reads the preference type score associated with the product purchased by the customer from the product information, uses each read preference category score instead of the customer's preference type score, and processes Step S160 and Step S165. You may do so. The client terminal 6 displays a recommendation screen based on the screen data received from the recommendation device 2. Alternatively, the selection unit 244 may transmit information such as the product name and price set in the recommended product information read in step S160 to the client terminal 6 as recommended information. The client terminal 6 transmits the received recommendation information to the POS system. The POS system may print recommended information on a receipt as recommended information, or may issue a coupon on which recommended information is printed for sales promotion.

Further, the web server 7 may transmit to the recommendation device 2 the customer ID of the customer and information on products purchased by the customer on a website such as an EC site provided by the web server 7. The processing unit 230 of the recommendation device 2 writes the product ID of the received product into the purchase history of the customer information specified by the customer ID. The selection unit 244 performs the same process as described above when information on a product purchased by a customer is acquired from the client terminal 6. The customer terminal 8 displays a recommendation screen based on the screen data received from the recommendation device 2.

In addition, the selection unit 244 of the recommendation device 2 refers to the customer information stored in the customer information storage device 3 and selects a product to be recommended among other customers whose preference type score is similar to the customer's preference type score. You may also calculate the percentage of customers who Similarity is determined by the inner product of the vectors represented by the taste type scores or the angle formed by the vectors represented by the taste type scores. The selection unit 244 generates screen data for a recommendation screen that displays the calculated ratio, and transmits it to the client terminal 6.

Note that the analysis information does not need to include information on the customer's attributes, nor does it need to include information on the customer's biological body. Further, the calculation of the preference type score based on the attribute information may be performed before the calculation of the preference type score based on the customer's answer.

Further, the calculation information of the score information by answer, the score information by attribute, and the score information by biometric information does not need to include information on the type of calculation. The calculation unit 242 may add up the scores of each preference category indicated by the calculation information, or may calculate a weighted average.

Note that the recommendation unit 240 of the recommendation device 2 may select products to be recommended using AI (artificial intelligence). For example, the recommendation unit 240 uses a recommended product selection model that inputs analysis information such as customer attributes, answers to each question, and biometric information, and outputs a recommendation score that quantitatively represents the degree of recommendation of each product. . The recommendation unit 240 inputs the customer attribute information acquired in step S105 and the answer and biometric information received in step S130 into the machine-learned recommended product selection model, and calculates a recommendation score for each product. get. The recommendation unit 240 selects a product that belongs to the product category input in step S110 and is handled by the client using the client terminal 6 (or another client affiliated with the client), The recommendation ranking of the selected product is determined in descending order of recommendation score. The recommended product selection model is obtained by machine learning using big data that associates customer analysis information with the products purchased by that customer as learning data.

Further, the calculation unit 242 may use a preference type determination model in which the customer's attributes, answers to each question, and biometric information are input and the score of each preference category is output to calculate the customer's preference type score. The calculation unit 242 inputs the customer attribute information that the information acquisition unit 241 acquired in step S105 and the answer and biometric information received in step S130 into the machine-learned preference determination model, and calculates the score of each preference category. obtain. The calculation unit 242 uses the obtained score for each preference category as the customer's preference type score. The preference type determination model is obtained by machine learning using big data as learning data that associates customer analysis information with the customer's correct preference type score.

Further, the selection unit 244 may use a recommended product selection model that inputs the score of each preference category and outputs a recommendation score that quantitatively represents the degree of recommendation of each product. The selection unit 244 inputs the score of each preference category represented by the customer's preference type score into a machine-learned recommended product selection model to obtain a recommendation score for each product. The recommended product selection model is obtained through machine learning using big data as learning data that associates a customer's preference type score with the products purchased by that customer.

FIG. 9 is a diagram showing a display example of a question display screen displayed on the client terminal 6 or the customer terminal 8. Here, the case where the first to third questions are question numbers 1 to 3 will be explained as an example. The template TM1 used for the question display screen G1 for the first question (n=1) includes a display G11 indicating that it is the nth question, a question content G12, a play button G13 for instructing playback of audio data, and a YES button G14 for inputting an answer of , and button G15 for inputting a NO answer. Audio data of music A, which is preference determination material, is embedded in the play button G13. The information acquisition unit 241 generates screen data of the question display screen G1 using the value "1" for n in the display G11.

The template TM2 used for the question display screen G2 for the second question (n=2) includes a display G21 indicating that it is the nth question, a question content G22 asking whether you like the phrase, and a button for inputting a YES answer. G23 and a button G24 for inputting a NO answer. Text data of phrase A, which is preference determination material, is embedded in the question content G22. The information acquisition unit 241 generates screen data for the question display screen G2 using the value "2" for n in the display G21.

The template TM3 used for the question display screen G3 for the third question (n=3) includes a display G31 indicating that it is the nth question, a question content G32, a button G33 for selecting the first image, and a second image. It includes a button G34 for selecting the second image. Image data of images B and C, which are taste determination materials, are embedded in buttons G33 and G34, respectively. The information acquisition unit 241 generates screen data of the question display screen G3 using the value "3" for n in the display G31.

FIG. 10 is a diagram showing a display example of the preference type display screen G4 displayed on the client terminal 6 or customer terminal 8. The taste type display screen G4 includes a taste type map G41, a customer's taste type score G42, and a pie chart G43 representing the ratio of taste categories. The taste type map G41 shows regions classified by taste type in a two-dimensional map. Each taste type is represented by a fixed coordinate range in the taste category space. Points P corresponding to customer preference type scores are plotted on the preference type map G41. A pie chart G43 shows the ratio of scores for each preference category that constitutes the customer's preference type score. Note that the analysis result output unit 243 may include an explanatory text according to the preference type score on the preference type display screen G4.

The first learning unit 251 of the recommendation device 2 uses the customer's preference type score calculated in the process of FIG. Update the values of parameters used in the selection algorithm for selecting products to recommend based on the preference categories of. For example, the first learning unit 251 totals the preference type scores of purchasing customers who have purchased a certain product for each preference category, and selects the preference category with the highest total score. If the selected preference category is different from the preference category of the product purchased by the customer, the first learning unit 251 calculates the preference type score corresponding to the product using the purchasing customer's analytical information. Change the value of calculation information so that the score of the category becomes high. Alternatively, when the first learning unit 251 calculates the preference type score using the analysis information of the customer, the ratio of preference categories in the calculated preference type score is the preference type score of all products purchased by the customer. The value of the calculation information is changed so that it approaches the ratio of preference categories in the average of .

Further, the second learning unit 252 of the recommendation device 2 determines the preference category to which the product corresponds based on the customer's preference type score calculated in the process of FIG. 8 and the preference category corresponding to the product purchased by the customer. or change the preference type score. For example, the second learning unit 252 totals the preference type scores of purchasing customers who have purchased a certain product by preference category, and rewrites the preference category to which the product corresponds to the preference category with the highest total score. Alternatively, the second learning unit 252 calculates the average of the purchasing customers' taste type scores, and changes the score of each taste category indicated by the taste type score corresponding to the product so as to approach the calculated average.

Alternatively, the learning unit 250 of the recommendation device 2 may use the trained recommended product selection model, preference type determination model, and recommended product selection model based on the analysis information and preference type score obtained in the process of FIG. It may be updated using a preference category or a preference type score corresponding to the product purchased by the customer, or the product purchased by the customer.

As described above, the recommendation device 2 asks customers questions regarding their preferences and intentions. For example, the recommendation device 2 presents a plurality of images, keywords, sounds (music), etc. to the customer, and asks the customer to decide whether they like or dislike each of them. By asking questions about items, colors (patterns), sounds, and other sensibilities, it is possible to obtain, for example, a preference type score representing a customer's lifestyle tendency. For example, if the answer is that they like woodgrain images, the score for the preference category "natural" will increase.

The recommendation device 2 determines and outputs the customer's preference type expressed as a combination of preference categories based on the customer's response results. By asking a plurality of questions related to various lifestyles, the recommendation device 2 can indicate lifestyle trends (also referred to as personas) for each customer in terms of preference category ratios. The recommendation device 2 receives the designation of a product or service classification, and recommends products or services that match the customer's lifestyle trends from a database that links products and services in that classification with preference categories. . Based on the knowledge that psychologically, people's preferences have a certain direction (such as liking a certain color or design), it is possible to estimate the products a customer likes based on their preference type. Since products from various categories and fields can be associated with the same axis of lifestyle, even if the product is in a different category or field than the customer is consciously searching for, it will still be a product that fits the customer's lifestyle. It is also possible to make recommendations.

Further, the recommendation device 2 can also be used as a recommendation engine linked from a client's website. The recommendation device 2 selects a recommendation target from a group of products and services that the client can provide.

The recommendation device 2 may use the addition or weighted average of the scores of each preference category associated with the preference judgment material as an algorithm for calculating the customer's preference type score from the answer results to the questions, and may also use machine learning. May be used. Furthermore, the recommendation device 2 can improve the accuracy of product recommendations by using attribute information such as gender, age, and occupation in addition to preference determination materials. Then, the taste type score may be updated by taking the product purchased by the customer at an actual store as the material for determining preference, and assuming that the customer "likes" the product. Stores can suggest products and services that match the customer's preferences from among a group of products and services that they want to encourage the customer to purchase.

According to the embodiment described above, it is possible to propose products, services, etc. to a customer without depending on the customer's purchase history. It becomes possible to propose products that match the tastes of customers.

FIG. 11 is a diagram showing an example of a business format to which the recommendation system 1 is applied.
(1) Customers search for products such as merchandise sales, food and beverages, and travel.
(2) Customers visit stores such as specialty stores, showrooms (SR), mass retailers, shopping centers, and model rooms to search for products. A client terminal 6 is installed in the store. The client terminal 6 may be a terminal for introducing customer service, or may be a signage. Alternatively, the customer accesses a company site or mail order site provided by the web server 7 from the customer terminal 8. Here, we will explain a case where two customers, a husband and wife, visit a nearby furniture specialty store with the purpose of purchasing a new set of furniture and changing the interior image. The customer receives a recommendation from the client terminal 6 at the recommendation of a dedicated coordinator.

(3) The customer selects the product category such as the desired product or service to be recommended. The client terminal 6 accesses the recommendation device 2 and transmits the product classification representing the interior product (furniture) selected by the customer.
(4) The customer inputs the customer's attributes, such as being a married couple in their 30s (no children), both working, living in Tokyo, and renting an apartment, on the screen displayed by the client terminal 6. The client terminal 6 transmits the input customer attributes to the recommendation device 2.
(5) The client terminal 6 displays the question screen received from the recommendation device 2. The customer intuitively inputs answers on the question screen displayed on the client terminal 6. For example, an image or the like is displayed randomly for a short time on the client terminal 6, and the customer inputs whether or not he likes the image (not only his preference but also his interest, etc.).
(6) The client terminal 6 senses biometric information and transmits it to the recommendation device 2. If the product classification is cosmetics or the like, the client terminal 6 may transmit the results of skin color measurement etc. to the recommendation device 2.

(7) The recommendation device 2 determines the customer's lifestyle tendency (preference type) using the customer's attributes, question answers, and biometric information based on an algorithm applying deep psychology.
(8) The client terminal 6 displays a preference type display screen showing the lifestyle trend determination results. The coordinator provides an explanation to the customer based on the presented preference type display screen. When presenting the preference type display screen to the customer terminal 8, the recommendation device 2 presents the preference type display screen including explanatory text and the like.

(9) The recommendation device 2 selects products that match the customer's lifestyle trends based on prior matching of lifestyle and products. For example, products to be recommended can be narrowed down based on the product grade desired by the customer. Further, the recommendation device 2 can select and present a recommendation target according to the ratio of the customer's preference categories by performing AI analysis of preference type scores representing lifestyle trends.

(10) The client terminal 6 presents information on the recommended product selected by the recommendation device 2. The recommendation device 2 may present the taste type score and the matching tendency of the recommended product to the client terminal 6.

(11) At the store, the coordinator explains the products while looking at the information displayed on the client terminal 6. In addition, in the case of a couple, the recommendation may be made for each person, or the recommendation may be made for one of them who has the right to make the decision. The customer decides on the product to purchase through dialogue with the coordinator.
(12) When the customer decides to purchase, the client notifies the recommendation device 2 from the client terminal 6 of information about the product purchased by the customer.

Note that the computer reads programs for realizing the functions of the above-mentioned recommendation device 2, customer information storage device 3, preference judgment material storage device 4, product information providing device 5, client terminal 6, web server 7, and customer terminal 8. The above-mentioned recommendation device 2, customer information storage device 3, taste judgment material storage device 4, and product information provision are performed by recording the program on a possible recording medium, and loading the program recorded on this recording medium into a computer system and executing it. Processing may be performed as the device 5, client terminal 6, web server 7, and customer terminal 8. Here, "reading a program recorded on a recording medium into a computer system and executing it" includes installing the program on the computer system. The "computer system" here includes hardware such as an OS and peripheral devices. Further, a "computer system" may include a plurality of computer devices connected via a network including the Internet, a WAN, a LAN, a communication line such as a dedicated line, etc. Furthermore, the term "computer-readable recording medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs, and CD-ROMs, and storage devices such as hard disks built into computer systems. In this way, the recording medium storing the program may be a non-transitory recording medium such as a CD-ROM. The recording medium also includes a recording medium provided internally or externally that can be accessed from the distribution server to distribute the program. The program code stored in the recording medium of the distribution server is executed by the recommendation device 2, the customer information storage device 3, the preference judgment material storage device 4, the product information providing device 5, the client terminal 6, the web server 7, and the customer terminal 8. The code may be different from the program code in the possible format. That is, it is downloaded from the distribution server and installed in a format that can be executed by the recommendation device 2, customer information storage device 3, preference judgment material storage device 4, product information providing device 5, client terminal 6, web server 7, and customer terminal 8. As long as it is possible, the format in which it is stored on the distribution server does not matter. Note that after the program is divided into multiple parts and downloaded at different timings, the recommendation device 2, the customer information storage device 3, the preference judgment material storage device 4, the product information providing device 5, the client terminal 6, the web server 7, and the customer terminal 8 The configurations that are combined and the distribution servers that distribute each of the divided programs may be different. Furthermore, a ``computer-readable recording medium'' refers to a storage medium that retains a program for a certain period of time, such as volatile memory (RAM) inside a computer system that serves as a server or client when a program is transmitted via a network. This shall also include things. Moreover, the above-mentioned program may be for realizing a part of the above-mentioned functions. Furthermore, it may be a so-called difference file (difference program) that can realize the above-mentioned functions in combination with a program already recorded in the computer system.

1... Recommendation system, 2... Recommendation device, 3... Customer information storage device, 4... Preference judgment material storage device, 5... Product information providing device, 6... Client terminal, 7... Web server, 8... Customer terminal, 210... Communication 211... Receiving section, 212... Transmitting section, 220... Storage section, 230... Processing section, 240... Recommendation section, 241... Information acquisition section, 242... Calculation section, 243... Analysis result output section, 244... Selection section, 250...Learning part, 251...First learning part, 252...Second learning part

Claims (9)

an information acquisition unit that acquires analysis information that is one or more pieces of information regarding at least one of customer preferences and intentions;
a calculation unit that calculates a preference type score that quantitatively represents the degree to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the analysis information acquired by the information acquisition unit; ,
Selecting the product to be recommended to the customer based on the preference type score calculated by the calculation unit and product information including at least preference category information indicating the preference category to which each of a plurality of products corresponds, and selecting the selected product. a selection section that outputs information about the product;
Based on the preference type score of the customer calculated by the calculation unit and the preference category information of the product selected by the customer, the selection unit makes a recommendation based on the customer's preference type score and the preference category information. a first learning unit that changes the value of a parameter used in an algorithm for selecting the product;
A recommendation device comprising: an information acquisition unit that acquires analysis information that is one or more pieces of information regarding at least one of customer preferences and intentions;
a calculation unit that calculates a preference type score that quantitatively represents the degree to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the analysis information acquired by the information acquisition unit; ,
Selecting the product to be recommended to the customer based on the preference type score calculated by the calculation unit and product information including at least preference category information indicating the preference category to which each of a plurality of products corresponds, and selecting the selected product. a selection section that outputs information about the product;
a second learning unit that changes the preference category information based on the preference type score of the customer calculated by the calculation unit and the product selected by the customer;
A recommendation device comprising: The information acquisition unit further acquires analysis information that is information regarding attributes or biometric information of the customer.
The recommendation device according to claim 1 or claim 2, characterized in that: The product information includes information on classification of the product,
The selection unit selects the product to be recommended to the customer from among the products in the category selected by the customer.
The recommendation device according to any one of claims 1 to 3 , characterized in that: The selection unit selects the product to be recommended to the customer from among the products in a classification different from the classification selected by the customer.
5. The recommendation device according to claim 4 . The preference category information is represented by a value that quantitatively represents the extent to which each of the plurality of preference categories matches the product,
The recommendation device according to any one of claims 1 to 5 . The information acquisition unit presents a question to the customer to obtain the analytical information regarding at least one of the customer's preferences and intentions as an answer.
The recommendation device according to any one of claims 1 to 6 , characterized in that: A recommendation method executed by a recommendation device,
an information acquisition step of acquiring analysis information that is one or more information regarding at least one of customer preferences and orientation;
a calculation step of calculating a preference type score that quantitatively represents the extent to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the analysis information obtained in the information acquisition step; ,
a selection step of selecting the product to be recommended to the customer based on the preference type score calculated in the calculation step and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds; ,
an output step of outputting information regarding the product selected in the selection step;
has
Based on the preference type score of the customer calculated in the calculation step and the preference category information of the product selected by the customer, in the selection step, the first learning unit calculates the customer's preference type score and changing the value of a parameter used in an algorithm for selecting the product to recommend based on the preference category information;
A recommendation method characterized by the following . A recommendation method executed by a recommendation device,
an information acquisition step of acquiring analysis information that is one or more information regarding at least one of customer preferences and orientation;
a calculation step of calculating a preference type score that quantitatively represents the extent to which each of a plurality of preference categories used as elements representing preference trends matches the customer, based on the analysis information obtained in the information acquisition step; ,
a selection step of selecting the product to be recommended to the customer based on the preference type score calculated in the calculation step and product information including at least preference category information indicating the preference category to which each of the plurality of products corresponds; ,
a selection step of outputting information regarding the product selected in the selection step;
has
a second learning unit changes the preference category information based on the preference type score of the customer calculated in the calculation step and the product selected by the customer;
A recommendation method characterized by the following.

Images