JP2010055391A - Merchandise analyzing device, merchandise offering device, and merchandise analysis method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically analyze sales, a trend or the like by automatically extracting features of merchandise from a merchandise image. <P>SOLUTION: At least one of merchandise color, shape and pattern is extracted from an merchandise image as a keyword tag or feature quantity. The extracted keyword tag or feature quantity is stored in association with merchandise information including the number of sales or a price and stored into a merchandise database. A user database stores user's purchase history, and users similar in preference are extracted as a user group on the basis of the purchase history. Merchandise purchased by the user group is extracted from among merchandise stored in the merchandise database. A trend value for the extracted merchandise is calculated on the basis of a keyword tag or the feature quantity and the merchandise information. Merchandise is ranked on the basis of the trend value. Merchandise at the top is displayed on a screen to thereby offer proper merchandise which encourages users to buy it. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a product analysis device, a product proposal device, a product analysis method, and a program, and more particularly to a product analysis device, a product proposal device, a product analysis method, and a program that perform analysis based on image characteristics.

  Conventionally, there have been many needs for grasping trends such as which products are sold and how much. Patent Document 1 is one of the methods for satisfying the needs, and discloses the content of attaching an IC tag to a product and managing the sales in the store. Japanese Patent Application Laid-Open No. 2004-228561 discloses that information is easily managed by extracting a keyword or pattern from a search target and performing a search based on the extracted keyword or the like.

In addition, on the Internet site, there are many techniques for presenting high-selling products to consumers by using the above-described method and leading to further purchase.
JP 2001-183051 A JP 2004-145586 A

  However, simply searching and managing the sales of each product using the methods described in Patent Documents 1 and 2 will increase or decrease the sales for reasons specific to that product such as a discount amount. Therefore, it is not possible to grasp the trend of the best selling in a certain field.

  For example, consider a case in which products with high sales are presented to consumers based on the sales of each product. It is common for a big hit product to be sold out, but there is no point in proposing a sold-out product, so it is necessary to propose another product to lead to further purchases. However, if a hit product is simply popular and is not attracting attention as a whole product of the same kind, no matter how much other similar products are proposed, it will not lead to purchase.

  Therefore, in order to examine a product that leads to purchase, that is, a group of products that are sold, it is necessary to examine the sales for each category and feature of each product. However, it is difficult to maintain a certain rule when the administrator determines the category or feature of each product. Specific examples are shown below.

  FIG. 13 shows a problem relating to the product name. FIG. 13A is an example of a product named “Knit Parka Dress”, FIG. 13B is an example of a product named “Boatneck Cocoon Tunic”, and FIG. This is an example of a product with the name “long knit”.

  The three products shown in FIGS. 13A to 13C are similar to each other, but there is no common point in the product names. In particular, the end of the product name often indicates the category of the product, but when these three products are classified into categories from the product name, they all belong to different categories even though they have similar shapes. turn into.

  FIG. 14 is a problem relating to color. FIG. 14A shows an example of a product tagged with a color “turquoise”, and FIGS. 14B and 14C show an example of a product tagged with a color “blue”.

  According to FIG. 14, it is clear that the color of the product shown in (a) is similar to the color of the product shown in (b), but when looking at the color manually attached to the administrator as a tag, The product in the center will be the same color as the product on the right.

  FIG. 14 shows an example of a single color, but it becomes more complicated when there are multiple colors. An example in the case where a product includes a plurality of colors is shown in FIG. FIG. 15A shows a white sweater including a white, light gray, and light pink three-color lattice pattern, and FIG. 15B includes a dark gray, black, and dark pink three-color lattice pattern. A dark gray sweater.

  In each of FIGS. 15A and 15B, pink includes a lattice pattern, but the product shown in (a) is tagged with “white”, which has a large proportion of the entire product, whereas in (b) The product shown has a tag “pink” which is a characteristic color even though the area is small. That is, the same color tag is attached to a sweater having the same background color but the same lattice pattern, or a different color tag is attached to a sweater having the same background color.

  Such differences do not arise from differences in manufacturers or brands. Even in the same brand, even in the case of terrible products, there are differences in the way names and tags are attached. That is, the difference between the names and tags as described above is caused by manually attaching the names and tags. Therefore, when manually attaching the names and tags, it is not possible to propose appropriate similar products. . This is particularly a problem in areas where appearance is important (apparel, interior, etc.).

  The present invention has been made in view of such circumstances, a product analysis device, a product proposal device, which can automatically analyze sales, trends, etc. by automatically extracting features of the product from a product image, The object is to provide a product analysis method and program.

  The product analysis apparatus according to claim 1, an acquisition unit that acquires an image of a product and sales information of the product, analyzes an image of the product acquired by the acquisition unit, and calculates a feature amount of the product. Calculated by means for calculating and / or means for creating a keyword tag indicating the characteristics of the product, sales information acquired by the acquisition means, and means for calculating the feature quantity from an image of the product corresponding to the sales information A product database that stores the associated feature quantity and / or the keyword tag created by the keyword tag creation means in association with each other, and a trend value calculation means that computes a trend value for each feature quantity and / or keyword tag The trend value is a value based on the sum of sales information of all products for which the same feature amount is calculated and / or the same keyword tag is created. Trend value calculating means for calculating, ranking means for ranking the feature quantities and / or keyword tags based on the trend value calculated by the trend value calculating means, and ranking performed by the ranking means Output means for outputting the result of the attachment.

  According to the product analysis apparatus of claim 1, the image of the acquired product is analyzed, and the feature tag of the product is calculated and / or the keyword tag indicating the feature of the product is created, and the product is sold. The information is stored in the product database in association with the feature amount calculated from the product image and / or the created keyword tag. A feature value and / or keyword is calculated for each feature value and / or keyword tag as a trend value based on the total sales information of all products for which the same feature value is calculated and / or the same keyword tag is created. Based on the trend value calculated for each tag, ranking of feature quantities and / or keyword tags is performed, and the result is output. Thereby, it is possible to calculate a feature amount and / or create a keyword tag based on a predetermined standard without variation. In addition, since there is no variation in the criteria for calculating feature amounts and / or creating keyword tags, it is possible to accurately extract products having similar features (similar products). Also, by ranking the feature quantities and / or keyword tags, ranking can be performed for each product (similar product) having similar characteristics, and the trend can be grasped.

  The product analysis apparatus according to claim 2 is characterized in that, in the product analysis apparatus according to claim 1, a feature amount included in the specific information is extracted from the product database based on specific information for which trend analysis is desired. A first extraction means for extracting the keyword tag included in the specific information from the product database as a keyword tag group, and the trend value calculation means is extracted by the first extraction means. The feature amount and / or the trend value of the keyword tag included in the feature amount group and / or the keyword tag group is calculated.

  According to the product analysis apparatus according to claim 2, based on specific information (for example, color, shape, pattern, etc.) for which trend analysis is desired, a feature amount included in the specific information is stored in the feature amount group from the product database. And / or keyword tags included in specific information from the product database are extracted as keyword tag groups. For example, when it is desired to perform trend analysis using colors, feature quantities including color information are extracted as feature quantity groups from the product database, and / or keyword tags including color information are extracted as keyword tag groups. Then, the feature value and / or keyword tag trend value included in the extracted feature value group and / or keyword tag group is calculated, and the feature value and / or keyword tag ranking is calculated based on the calculated trend value. Do. Thereby, a true trend can be grasped.

  The product analysis device according to claim 3 is the product analysis device according to claim 1 or 2, wherein the trend value calculation means calculates a trend value for each product based on the sales information, and the ranking is performed. The means ranks the products for each feature quantity and / or keyword tag for which the ranking has been performed based on the trend value for each product calculated by the trend value calculation means. .

  According to the product analysis apparatus according to claim 3, the trend value is calculated for each product based on the sales information, and the product ranking is performed for each feature amount and / or keyword tag based on the calculated trend value. I do. Thereby, ranking of goods can be performed automatically. In addition, since the ranking of products is performed for each feature amount and / or keyword tag, it is possible to grasp a more detailed trend.

  The product analysis device according to claim 4 is the product analysis device according to any one of claims 1 to 3, wherein the means for calculating the feature amount is a color included in an image of the product acquired by the acquisition unit. And the means for creating the keyword tag creates the text representing the most contained color as the keyword tag. Features.

  According to the product analysis apparatus according to claim 4, the most contained color among the colors included in the product image is calculated as a feature amount, and the most contained color among the colors included in the product image is represented. Create text as keyword tags. Thereby, the color of goods can be used as a feature-value and / or a keyword tag.

  The product analysis device according to claim 5 is the product analysis device according to any one of claims 1 to 3, wherein the means for creating the keyword tag is based on pattern recognition from an image of the product acquired by the acquisition unit. At least one of the shape and pattern of the product is extracted, and text representing at least one of the extracted shape and pattern of the product is created as the keyword tag.

  According to the product analysis apparatus of claim 5, at least one of the shape and pattern of the product is extracted from the image of the product by pattern recognition, and text representing at least one of the extracted shape and pattern is created as a keyword tag. To do. Thereby, the shape and pattern of goods can be used as a keyword tag.

  The product proposal device according to claim 6 is calculated by the product analysis device according to any one of claims 1 to 5, a product purchased by a user, and a unit that calculates the feature amount from an image of the product. Based on the purchase history stored in the user database and the user database that stores the feature quantity and / or the purchase history including the keyword tag created by the keyword tag creation unit in association with each other, the hobbies and preferences are similar. User group extraction means for extracting a user as a user group, feature quantities included in the purchase history of the user group extracted by the user group extraction means from the product database as feature quantity groups, and / or the product database The keyword tag included in the purchase history of the user group extracted by the user group extraction means from the keyword tag group And extracting the product for which the feature amount included in the purchase history of the user group is calculated and / or the product for which the keyword tag included in the purchase history of the user group is created from the product database. Extraction means, and the trend value calculation means calculates the feature value and / or the trend value of the keyword tag included in the feature value group and / or the keyword tag group extracted by the second extraction means, and The trend value of the product extracted by the second extraction unit is calculated, and the ranking unit includes a feature amount and / or a keyword included in the feature amount group and / or the keyword tag group extracted by the second extraction unit. The feature quantity and / or the keyword tag for which the ranking of the tags is performed and the ranking of the products extracted by the second extraction unit is performed. And the output means ranks the feature quantities and / or keyword tags in the upper predetermined rank among the feature quantities and / or keyword tags ranked by the ranking means by the ranking means. Among the products for which the product has been executed, products in a predetermined order are acquired from the product database, and the acquired products are output as products proposed to the user.

  According to the product proposal device of claim 6, users having similar hobbies and preferences are extracted as a user group based on the purchase history stored in the user database, and the feature amount included in the purchase history of the user group from the product database And / or a keyword tag included in the purchase history of the user group from the product database as a keyword tag group, and the extracted feature amount and / or the feature amount included in the keyword tag group and / Or calculate trend values for keyword tags. In addition, the product for which the feature amount included in the purchase history of the user group is calculated and / or the product for which the keyword tag included in the purchase history of the user group is created is extracted, and the ranking of the extracted product is ranked. Is performed for each feature amount and / or keyword tag. Then, for the feature quantity and / or keyword tag that has been ranked, the top-ranked feature quantity and / or keyword tag, the top-ranked product among the ranked products Is acquired from the product database, and the acquired product is output as a product proposed to the user. Thereby, ranking of an index and goods can be performed for every user group. Further, since the output product is determined for each user group, it is possible to propose an appropriate product that matches the current trend and induces the user's willingness to purchase.

  The product proposal device according to claim 7 is the product proposal device according to claim 6, wherein the user purchase history acquisition unit acquires a purchase history of a predetermined user, and the purchase history acquired by the user purchase history acquisition unit And a user group designating unit for designating a user group to which the predetermined user belongs from among the user groups extracted by the user group extracting unit based on the purchase history stored in the user database; A feature group and / or keyword tag group of the user group designated by the user group designating means is extracted from the feature quantity group and / or keyword tag group extracted by the extracting means, and the designated user group is extracted. A product for which a feature value included in the feature value group is calculated and / or a keyword tag included in the keyword tag group of the specified user group is created. A third extraction unit for extracting the product, and the trend value calculation unit includes the feature amount and / or the keyword tag included in the feature amount group and / or the keyword tag group extracted by the third extraction unit. And the trend value of the product extracted by the third extraction means is calculated, and the ranking means adds the feature value group and / or the keyword tag group extracted by the third extraction means. A ranking of feature quantities and / or keyword tags included is performed, and ranking of products extracted by the third extraction means is performed for each feature quantity and / or keyword tag for which ranking has been performed. And

  According to the product suggestion apparatus of the seventh aspect, the user group to which the predetermined user belongs is specified from the user group based on the purchase history of the predetermined user and the purchase history stored in the user database. The feature amount included in the purchase history of the user group to which the predetermined user belongs is extracted as a feature amount group from the product database, and / or the keyword tag included in the purchase history of the user group to which the predetermined user belongs from the product database is the keyword tag. A feature value and / or a trend value of a keyword tag included in the extracted feature value group and / or keyword tag group is calculated as a group. In addition, the product in which the feature amount included in the purchase history of the user group to which the predetermined user belongs and / or the product in which the keyword tag included in the purchase history of the user group is created are extracted, and the rank of the extracted product For each feature amount and / or keyword tag for which ranking is performed. Then, for the feature quantity and / or keyword tag that has been ranked, the top-ranked feature quantity and / or keyword tag, the top-ranked product among the ranked products Is acquired from the product database, and the acquired product is output as a product proposed to the user. Thereby, it is possible to propose a product that matches a hobby preference of a specific user and is in line with the trend. Therefore, it is possible to propose a product that will motivate the user to purchase.

  The product proposal device according to claim 8 is the product proposal device according to claim 7, wherein the feature quantity ranked by the ranking unit and / or a predetermined number of feature quantities among the keyword tags. And / or a product in a higher order among the products associated with the keyword tag is acquired from the product database, and the acquired product is obtained based on the purchase history of the predetermined user acquired by the user purchase history acquisition means. An unpurchased product extracting unit that extracts products not included in the purchase history of the predetermined user from among the items, and the output unit outputs the product extracted by the unpurchased product extracting unit as a product to be proposed to the user It is characterized by doing.

  According to the product proposal device according to claim 8, the higher-ordered feature quantity and / or keyword tag among the higher-order predetermined number of feature quantities and / or the product associated with the keyword tag Based on the purchase history of a predetermined user, products not included in the purchase history of the predetermined user are extracted from the products in a predetermined order, and the product is output as a product proposed to the user. As a result, it is possible to propose a product that will motivate the user to purchase.

  The product analysis method according to claim 9 includes a step of acquiring a product image and sales information of the product, calculating the feature amount of the product by analyzing the acquired product image, and / or Alternatively, a feature tag calculated by the step of creating a keyword tag indicating the feature of the product, the step of calculating the feature value from the acquired sales information, and an image of the product corresponding to the sales information and / or the A step of associating the keyword tag created in the step of creating a keyword tag with the keyword tag and storing it in a product database; and a step of calculating a trend value for each feature amount and / or keyword tag, wherein the same feature amount is calculated. And / or calculating a trend value based on the sum of sales information of all products for which the same keyword tag is created; Based on the trend value the calculated and performing ranking of said characteristic quantity and / or keyword tag, characterized in that it comprises the steps of: outputting a result of the ranking.

  According to a tenth aspect of the present invention, there is provided a program for causing a computing device to execute the commodity analysis method according to the ninth aspect.

  According to the present invention, it is possible to automatically analyze sales and trends by automatically extracting features of the product from the product image.

<First Embodiment>
FIG. 1 is a schematic diagram of the overall structure of a trend analysis apparatus 1 according to the first embodiment. The trend analysis apparatus 1 mainly includes a control unit 10, an input / output unit 11, a memory control unit 12, a product DB (database) 13, a DB (database) management unit 14, an image management unit 15, a decoder / The encoder 16, the feature amount extraction unit 17, the trend value calculation unit 18, and the output product determination unit 19 are configured.

  The control unit 10 functions as a control unit that performs overall control of the overall operation of the trend analysis apparatus 1 and also functions as a calculation unit that performs various types of calculation processing. The control unit 10 also has a memory space. The memory space stores firmware, which is a control program executed by the control unit 10, various data necessary for control, and is used as a work area for the control unit 10. The

  The memory control unit 12 controls reading / writing of a memory (not shown).

  The product DB 13 includes a product image (hereinafter referred to as a product image), sales information such as the number of goods received, the number of inventory, the number of sales, the price of the product, the description of the product, the sales time, the product name, the brand The information is stored in association with the feature quantity or keyword tag extracted from the product image in the feature quantity extraction unit 17 (detailed later). The product image and product information are input to the product DB 13 via a network or the like (not shown). Various data stored in the product DB 13 is managed by the DB management unit 14 for reading and writing.

  The image management unit 15 is used as a temporary storage area for image data acquired from the DB 13. The image data temporarily stored in the image management unit 15 is decoded by the decoder / encoder 16 and input to the feature amount extraction unit 17.

  The feature amount extraction unit 17 extracts keyword tags and feature amounts from the image data input from the image management unit 15. As the keyword tag and the feature amount, the color, shape, pattern, etc. of the product can be used. The keyword tags and feature amounts calculated by the feature amount extraction unit 17 are stored in the product DB 13 via the control unit 10 and the DB management unit 14. A method of extracting feature quantities and keyword tags by the feature quantity extraction unit 17 will be described in detail later.

  The trend value calculation unit 18 determines a trend value for each product based on the feature amount or keyword tag extracted by the feature amount extraction unit 17 and the product information stored in the product DB 13 for each product. calculate. A method of calculating the trend value by the trend value calculation unit 18 will be described in detail later.

  The output product determination unit 19 determines a product to be output from the input / output unit 11 based on the trend value calculated by the trend value calculation unit 18. The method of determining the product output by the output product determination unit 24 will be described in detail later. The output product determined by the output product determination unit 19 is output from the input / output unit 11.

  The operation of the trend analysis apparatus 1 configured as described above will be described. First, the flow of processing for registering a product in the product DB 13 will be described with reference to FIG.

  First, the control unit 10 acquires a product image from the product DB 13 via the DB management unit 14 and outputs the product image to the image management unit 15. The image management unit 15 outputs the image data decoded by the decoder / encoder 16 to the feature amount extraction unit 17, and the feature amount extraction unit 17 reads the input image data (step S10).

  The feature amount extraction unit 17 extracts keyword tags and feature amounts from the input image data (step S11). The feature amount extraction unit 17 extracts a representative color / shape / pattern of a product from the product image as a keyword tag or a feature amount as follows.

  The feature quantity extraction unit 17 removes the background from the product image, and extracts the color most contained in the image data from which the background has been removed as a representative color. The color is usually expressed in a three-dimensional RGB space, but as shown in FIG. 3, the color is expressed in one dimension in the present embodiment. The feature quantity extraction unit 17 creates a one-dimensional histogram as follows.

  First, the image data from which the background has been removed, that is, the RGB value of each pixel included in the clothing region is acquired for each pixel, and the acquired RGB value for each pixel is converted into an HSV value. HSV is a color space in which colors are represented by H (Hue: Hue), S (Saturation: Saturation), and V (Value: Lightness).

  Next, the HSV space is divided into predetermined color areas, the number of pixels in each color area is counted, and the result is formed into a histogram.

  Then, a weighted histogram is generated using the function shown below. In Equation 1, A, B, and C are constants, f (h, s, v) is a histogram before weighting, and f ′ (h, s, v) is a histogram after weighting.

  As a result, a histogram as shown in FIG. 3 is created, and representative colors are extracted from this histogram. The following two methods can be considered as the representative color extraction method, but the representative color may be extracted by any method.

  Method 1: Among the maximum values of f ′ (h, s, v), N colors having a large value are set as representative colors.

  Method 2: The color of the maximum value among the maximum values of f (h, s, v) is determined as the first representative color. When multiple representative colors are determined, the process of determining the next representative color as the maximum representative color excluding the approximate colors selected as representative colors is extracted until a desired number of representative colors are extracted. repeat.

  In the present embodiment, the color of the maximum value among the maximum values of f (h, s, v) is determined as the representative color. In the example shown in FIG. 3, since the frequency of the color 2 (shaded portion) is higher than the frequency of the color 1 (blacked portion), the color 2 is extracted as a representative color.

  The feature amount extraction unit 17 uses the absolute value of the representative color extracted in this way (the value on the horizontal axis of the histogram shown in FIG. 3) as the feature amount. Further, the feature amount extraction unit 17 creates a color name (for example, color 2) of the representative color extracted in this way as a keyword tag.

  The feature amount extraction unit 17 extracts the shape and pattern of the product by pattern matching the product image. Since various known techniques can be used for pattern matching, detailed description thereof is omitted. The feature amount extraction unit 17 creates the shape (for example, trapezoid skirt) or the pattern (for example, plain) extracted in this way as a keyword tag. Words that become keyword tags are stored in advance in the feature amount extraction unit 17 in association with patterns in a plurality of layers by an administrator or the like. Based on the extracted pattern, the feature quantity extraction unit 17 selects an optimum word for each layer from preset words, and creates the word as a keyword tag. For example, in the case of the product image shown in FIG. 3, the keyword tag “women's clothing” at the highest level, the keyword tag “skirt” at the second level, the keyword tag “skirt”, and the third keyword tag “ A trapezoidal skirt "is created.

  The feature amount extraction unit 17 outputs the feature amount and the keyword tag extracted in this way to the control unit 10. The control unit 10 outputs feature quantities and keyword tags to the product DB 13 via the DB management unit 14. The product DB 13 stores the input feature quantity or keyword tag in association with product information of the product corresponding to the product image from which the feature quantity or keyword tag is extracted (step S12). As a result, the product is registered in the product DB 13.

  The control unit 10 completes the registration of all the products, that is, extracts all the products for which product images and product information have been input to the product DB 13 (step S11) and stores them in the product DB 13. It is determined whether the registration process (step S12) is completed (step S13).

  If the product quantity and keyword tag extraction processing (step S11) and registration processing in the product DB 13 (step S12) have not been completed for all products for which product images and product information are input to the product DB 13, the product image Is returned to the step of reading (step S10), and the processing of steps S10 to S12 is performed again.

  When the feature value and keyword tag extraction processing (step S11) and the registration processing to the product DB 13 (step S12) have been completed for all products for which product images and product information have been input to the product DB 13, the processing ends. .

  Thereby, the process which registers goods in goods DB13 is complete | finished. Next, with reference to FIG. 4, a process for determining a process for determining an output product from among products registered in the product DB 13 will be described.

  The control unit 10 designates a product group for which a trend value is to be calculated from the products registered in the product DB 13 (step S14). The product group is specified by inputting a feature amount or a keyword tag to the control unit 10 via the input / output unit 11. In step S14, when the absolute value of the representative color that is the feature amount is input to the control unit 10, the control unit 10 outputs the feature amount to the feature amount extraction unit 17, and the feature amount extraction unit 17 receives the input. The keyword tag (for example, color 2) corresponding to the feature amount is output to the control unit 10. When the feature amount extraction unit 17 creates a keyword tag corresponding to a feature amount, if a predetermined range is provided for the feature amount, products having similar features (similar products) can be accurately extracted. .

  When a keyword tag is input to the control unit 10 from the feature amount extraction unit 17 or the input / output unit 11, a product with the keyword tag input from the products registered in the product DB 13 is specified as a product group. For example, when the keyword “women's clothing” is input to the control unit 10, the control unit 10 designates a product group to which the keyword tag “women's clothing” is attached from among the products registered in the product DB 13. To do.

  In step S14, if no feature value or keyword tag is input, that is, if no product group is specified, it is assumed that all products registered in the product DB 13 have been selected.

  The trend value calculation unit 18 calculates a trend value for each keyword tag based on the product information of the products registered in the product DB 13 for all the products belonging to the product group specified in Step S14 (Step S15). ). Trend values include, for example, (1) the absolute number of products sold within a certain period, (2) the total price of products sold within a certain period, and (3) the number of products sold and the number of arrivals within a certain period. (4) A combination of the above (1) to (3) (weighted addition of (1) to (3)) or the like can be used.

  For example, when a product group with a keyword tag “women's clothing” is specified in step S14, the trend value calculation unit 18 is one level lower than the keyword tag “women's clothing” (hereinafter referred to as the second item). The sum of the trend values is calculated for each keyword tag (for example, “enamel”, “bootie”, “race”, “bicolor”, “geometric pattern”, and “check”). When (1) the absolute number of products sold within a certain period is used as the trend value, the trend value calculation unit 18 has sold all the products for which the keyword tag of the second hierarchy has been created within the certain period. The total absolute number of products is calculated as the trend value.

  Also, trend values are calculated for all products in the product group. When (1) the absolute number of products sold within a certain period is used as the trend value, the absolute number of products sold within the certain period is set as the trend value of the product.

  The output product determination unit 19 determines the product to be output, the number of products, the output order, and the like based on the trend value calculated in step S15 (step S16). That is, the output product determination unit 19 determines a keyword tag that outputs a keyword tag having a trend value within a predetermined upper order, and outputs the determined keyword tag in descending order of the trend value. In addition, among the products for which the keyword tag determined to be output is created, it is determined as a product for which the trend value is within the upper predetermined rank, and the determined products are output in descending order of the trend value for each keyword tag. .

  For example, regarding the specific method of step S16, when the keyword tag “women's clothing” is input in step S14, and the trend value based on the absolute number of products sold within a certain period is calculated in step S15. The output product determination unit 19 outputs the keyword tags having the trend values of 1 to 3 in the order of the trend values of 1 to 3 in the product group to which the keyword tag of the second hierarchy is attached. . The output product determination unit 19 sets the trend value 1 as the product that outputs the products with the trend values 1 to 3 for each of the product groups to which the keyword tags with the trend values 1 to 3 are attached. Output in order of No. 3 to No. 3.

  Thereby, the process which determines an output goods from the goods registered into goods DB13 is complete | finished. The control unit 10 acquires a product image or product information determined as a product to be output from the output product determination unit 19 and inputs it to the input / output unit 11. The input / output unit 11 outputs this result to an external display means or the like.

  The output result is displayed as shown in FIG. This is a display example of a business management tool displayed on a store terminal or the like.

  As shown in FIG. 5, classification of the women's clothing is performed for each keyword tag in which the sum of the trend values of the keyword tags in the second hierarchy is No. 1 to No. 3. Then, for each keyword tag, a product name having a trend value of No. 1 to No. 3 is displayed. In this way, instead of just presenting the best-selling products, displaying the keyword tags together with the product information of the products with the keyword tags appropriately displays the sales, trends, popular products, etc. I can grasp it. In the example shown in FIG. 5, the product name is displayed because it is for business use, but a product image may be displayed.

  According to the present embodiment, since the feature of the product is automatically extracted from the product image, it is possible to accurately extract a product (similar product) having similar characteristics with no variation in the determination of the feature of the product. it can.

  Moreover, according to this Embodiment, since a similar product can be extracted exactly, the suitable similar product which matched the present trend can be proposed.

  In the present embodiment, processing such as product group specification (step S14) is performed using feature amounts and keyword tags extracted from the product image. However, if objective judgment is easy, automatic processing is performed. In addition to the feature quantity and keyword tag extracted automatically, a keyword tag input manually may be added. For example, in the case of clothes, the material (cotton, hemp, etc.) can be objectively determined regardless of individual subjectivity, but it is difficult to extract as a keyword tag from a product image. Therefore, similar items can be extracted more accurately by attaching keyword tags that are manually input for materials.

  In the present embodiment, the feature amount extraction unit 17 selects an optimal word for each hierarchy from preset words and sets the word as a keyword tag. Not limited to.

  For example, in addition to the keyword tag based on the shape, the keyword tag “plain color” based on the pattern or the keyword tag “beige” based on the color may be added, or in addition to the keyword tag based on the shape You may make it attach a feature-value. Further, a keyword tag based on a pattern or a keyword tag based on a color may be attached to a plurality of keyword tags.

  Further, when two or more features such as “trapezoid skirt” and “pleated skirt” are extracted from the same product image, a plurality of keyword tags in the same hierarchy may be attached.

  In the present embodiment, the output product determination unit 19 has been described as an example in which the output product is determined using the trend value based on the absolute number of products sold within a certain period and the keyword tag. The output product may be determined based on the trend value and product information (for example, product price, brand, etc.). When the output product is determined based on the trend value and the price of the product (for example, in ascending order of price), the product that matches the trend and has a low price can be output. Further, when the output product is determined based on the trend value and the brand, the product can be output not only in accordance with the trend but also in consideration of conditions specific to the brand.

<Second Embodiment>
In the second embodiment, a user's purchase history is acquired, and a product that matches the user's hobbies and preferences is proposed according to the purchase history. Hereinafter, the product recommendation device 2 according to the second embodiment will be described. The same parts as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof is omitted.

  FIG. 6 is a schematic diagram of the overall structure of the product recommendation device 2 according to the second embodiment. The product recommendation device 2 mainly includes a control unit 10, an input / output unit 11, a memory control unit 12, a product DB (database) 13, an image management unit 15, a decoder / encoder 16, and a feature amount extraction unit 17. And a trend value calculation unit 18, an output product determination unit 19, a user DB (database) 20, a DB (database) management unit 21, and a user group extraction unit 22.

  The user DB 20 stores information related to the user such as the user name, user ID, date of birth, and sex, and product information (hereinafter referred to as purchase history) of the product purchased by the user in association with each other. Information about the user and purchase history are input to the user DB 20 via a network (not shown).

  The DB (database) management unit 21 manages reading and writing of various data stored in the product DB 13 and the user DB 20.

  The user group extraction unit 22 extracts users having similar hobbies and preferences as users from the users stored in the user DB 20. As the user group, for example, a user group in which the ratio of a predetermined keyword tag included in the purchase history is equal to or greater than a predetermined threshold (such as a user group having many purchases of one-piece), or a user group having a characteristic amount within a predetermined range ( A group of users who prefer to purchase a predetermined color), a group of users whose average purchase price is within a predetermined range or above a predetermined threshold (such as a group of users who purchase a lot of high-end products), and a user whose age is within a predetermined range Group (such as a group of users who often purchase gal brands). The method by which the user group extraction unit 22 extracts a user group will be described in detail later.

  The operation of the product recommendation device 2 configured as described above will be described. The flow of processing for registering a product in the product DB 13 is the same as that in the first embodiment, so the description is omitted, and a product that matches the user's hobby preference is proposed from among the products registered in the product DB 13. Processing to determine as a product to be performed will be described with reference to FIG.

  Based on the purchase history stored in the user DB 20, for example, the feature amount, the keyword tag, the brand, the purchase price, etc., the user group extraction unit 22 receives information on the user and the purchase history stored in the user DB 20 from the user. Users with similar preferences are extracted as a user group (step S21). Information necessary for extracting a user group is performed by inputting a keyword tag and a feature amount to the control unit 10 via the input / output unit 11.

  The details of the method (step S21) in which the user group extraction unit 22 extracts a user group will be described by taking as an example the case of performing trend analysis on women's clothes. When the keyword “female” is input to the control unit 10 via the input / output unit 11, the control unit 10 extracts female users registered in the user DB as a female user group.

  Furthermore, high school students, university students, feminine OLs, and cool OLs are extracted as user groups in the hierarchy below the female user group. When the keyword “high school student” is input to the control unit 10 via the input / output unit 11, the control unit 10 extracts high school students from a group of female users from the date of birth registered in the user DB, This is a group of high school students.

  However, since college students and OLs have the same age group, extraction cannot be performed using the date of birth. Therefore, university students and OLs are extracted using the purchase history. For example, if the purchase history includes a college student-oriented brand in a predetermined ratio or more, it is estimated that the student is a college student. Therefore, the control unit 10 extracts this user as a user group of college students.

  For users who were not estimated to be university students, check whether the purchase history contains more “cool” or “feminine” keyword tags, and more “cool” keyword tags. A user is extracted as a cool OL user group, and a user including more “feminine” keyword tags is extracted as a feminine OL user group. The extraction of the cool OL and the feminine OL can be performed using a keyword tag or can be performed using a feature amount. For example, the user group extraction unit 22 extracts a user whose average feature value included in the purchase history is within a predetermined range including a color preferred by a cool user as a cool OL user group. A feminine OL user group is similarly extracted. Thereby, a user group with a similar hobby preference can be extracted automatically and accurately. In addition, when a user group is extracted using a feature amount, future processing is performed using a keyword tag corresponding to the input feature amount.

  The control unit 10 designates a user group whose trend value is to be calculated from the user group extracted in step S21 (step S22).

  The trend value calculation unit 18 calculates a trend value based on the product information of the products registered in the product DB 13 for all the products included in the user group specified in Step S22 (Step S23). Trend values include, for example, (1) the absolute number of products sold within a certain period, (2) the total price of products sold within a certain period, and (3) the number of products sold and the number of arrivals within a certain period. (4) A combination of the above (1) to (3) (weighted addition of (1) to (3)) or the like can be used.

  For example, when the user group of “high school students” is designated in step S22, the output product determination unit 19 acquires the purchase history of users included in the user group of “high school students” from the user DB 20, and stores the purchase history in the purchase history. The product group in which the included keyword tag is created is acquired from the product DB 13. For example, in the case where the keyword tag “enamel” is included in the purchase history and (1) the absolute number of products sold within a certain period is used as the trend value, the trend value calculation unit 18 selects “enamel”. The total of the absolute number of products sold within a certain period of all the products for which the keyword tag is created is calculated as a trend value.

  Also, trend values are calculated for all products in the product group. When (1) the absolute number of products sold within a certain period is used as the trend value, the absolute number of products sold within the certain period is set as the trend value of the product.

  The control unit 10 determines whether or not trend values have been calculated for all the user groups extracted in step S21 (step S24).

  If trend values have not been calculated for all user groups extracted in step S21 (NO in step S24), user group selection processing (step S22) and trend value calculation processing (step S23) are performed again. Judgment processing (step S24) is performed.

  When the trend value is calculated for all the user groups extracted in step S21 (YES in step S24), the output product determination unit 19 uses the trend value calculated in step S23. The product to be output, the number of products, the output order, etc. are determined for each user group extracted in step S21 (step S25). That is, the output product determination unit 19 determines, for each user group extracted in step S21, as a keyword tag that outputs a keyword tag having a trend value within the upper predetermined rank, and the determined keyword tag has a high trend value. Output in order. In addition, among the products for which the keyword tag determined to be output is created, it is determined as a product for which the trend value is within the upper predetermined rank, and the determined products are output in descending order of the trend value for each keyword tag. .

  For example, if a group of high school students, college students, feminine OLs, and cool OLs is extracted in step S21 and a trend value based on the absolute number of products sold within a certain period is calculated in step S23, the output is The merchandise deciding unit 19 outputs, for each user group, keyword tags whose trend values are No. 1 to No. 3 in the order of the trend values of No. 1 to No. 3. Then, the output product determination unit 19 outputs a product group with a trend value of No. 1 to No. 3 for a product group to which a keyword tag having a trend value of No. 1 is attached, that is, a product group that has been sold the most. Output in order of value 1 to 3. In addition, for each of the product groups to which the keyword tags having the trend values of No. 2 and No. 3 are attached, the products with the highest trend values, that is, the products that output the most sold products, that is, the products to be proposed.

  Thereby, the process which determines as a goods which proposes the goods which match a user's hobby preference is complete | finished. The control unit 10 acquires a product image or product information determined as a product to be output from the output product determination unit 19 and inputs it to the input / output unit 11. The input / output unit 11 outputs this result to an external display means or the like.

  The output result is displayed as shown in FIG. This is a display example for general users displayed on the top page of an EC (Electronic Commerce) site.

  As shown in FIG. 8, classification is performed for each user group “high school student”, “university student”, “feminine OL”, and “cool OL” in the second hierarchy, and the sum of trend values for each user group is No. 1 to Classification is performed for each keyword tag that is the third. For the keyword tag with the highest sum of trend values, the product image with the trend value of 1 to 3 is displayed, and for the other keyword tags, the product image with the highest trend value is displayed.

  In this way, instead of just presenting the best-selling product, as shown in FIG. 8, it is easier to grasp the trend by displaying the feature image and the keyword tag having a high trend value together with the product image. be able to. Moreover, since the product image is displayed, there is an advantage that it is easy to grasp the image.

  According to the present embodiment, since the output product is determined and displayed for each user group, it is possible to easily determine the trend. In addition, it is possible to propose an appropriate product that matches the current trend and induces the user's willingness to purchase.

<Third Embodiment>
The third embodiment proposes a product that matches a user's hobbies and preferences to a specific user. Hereinafter, the product recommendation device 3 according to the third embodiment will be described. In addition, about the part same as 1st Embodiment or 2nd Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  FIG. 9 is a schematic diagram of the overall structure of the product recommendation device 3 according to the third embodiment. The product recommendation device 3 mainly includes a control unit 10, an input / output unit 11, a memory control unit 12, a product DB (database) 13, an image management unit 15, a decoder / encoder 16, and a feature amount extraction unit 17. A trend value calculation unit 18, an output product determination unit 19, a user DB (database) 20, a DB (database) management unit 21, a user group extraction unit 22, and a communication I / F (interface) 31. Composed.

  The communication I / F 31 acquires information about the user and purchase history output from a user terminal (not shown), and outputs the acquired information to the control unit 10.

  The operation of the product recommendation device 3 configured as described above will be described. The flow of processing for registering a product in the product DB 13 is the same as that in the first embodiment, so the description is omitted, and a product that matches the user's hobby preference is proposed from among the products registered in the product DB 13. Processing to determine as a product to be performed will be described with reference to FIG.

  Based on the purchase history stored in the user DB 20, for example, the feature amount, the keyword tag, the brand, the purchase price, etc., the user group extraction unit 22 receives information on the user and the purchase history stored in the user DB 20 from the user. Users with similar preferences are extracted as a user group (step S21). High school students, university students, feminine OLs, and cool OLs are extracted as user groups below the user groups.

  The communication I / F 31 inputs information about the user and purchase history acquired from a user terminal (not shown) to the control unit 10, and the control unit 10 inputs information about the user and purchase history to the user group extraction unit 22 (step S31). ).

  The user group extraction unit 22 specifies the user group to which the user who acquired the information related to the user and the purchase history belongs from the user groups extracted in step S21 based on the information related to the user acquired in step S31 and the purchase history. (Step S32). For example, the user group extraction unit 22 compares the keyword tag or feature amount of each user group with the keyword tag or feature amount of the purchase history acquired in step S31, and is most similar to the keyword tag or feature amount of the purchase history. A user group having a high keyword tag or feature amount is designated as a user group to which the user belongs. The similarity of the keyword tag is represented, for example, by the ratio that the keyword tag of the purchase history is included in the keyword tag of each user group, and the similarity of the feature amount is, for example, the average value of the similarity of the purchase history and each user It is expressed as a ratio to the average value of the group similarity.

  The specific method of step S32 will be described by taking as an example the case where a user group of feminine, cool, sexy, casual, and natural is extracted in step S21. The user group extraction unit 22 compares the keyword tag or feature amount of the purchase history acquired in step S31 with the keyword tag or feature amount of each user group, and determines the keyword tag or feature amount of the purchase history acquired in step S31. Classify feminine, cool, sexy, casual and natural. As a result, the keyword tag or feature amount classified as feminine was 50%, the keyword tag or feature amount classified as cool was 0%, the keyword tag or feature amount classified as sexy was 30%, and classified as casual. When the keyword tag or feature amount is 0% and the keyword tag or feature amount classified naturally is 20%, the user group extraction unit 22 has the highest ratio of the keyword tag or feature amount classified as feminine. It is determined that this user belongs to the feminine user group because there are many.

  The trend value calculation unit 18 calculates a trend value based on the product information of the products registered in the product DB 13 for all the products included in the purchase history of the user group specified in step S32 (step S33). . Trend values include, for example, (1) the absolute number of products sold within a certain period, (2) the total price of products sold within a certain period, and (3) the number of products sold and the number of arrivals within a certain period. (4) A combination of the above (1) to (3) (weighted addition of (1) to (3)) or the like can be used.

  For example, when a feminine user group is specified in step S32 and a trend value is calculated based on the absolute number of products sold within a certain period in step S33, the output product determination unit 19 displays the feminine user group. The sum of trend values is calculated for each keyword tag included in the purchase history, for example, “enamel”, “bootie”, “race”, “pink”, and “polka dot pattern”.

  Based on the trend value calculated in step S33, the output product determination unit 19 determines the product to be output, the number of products, the output order, etc. for the user group specified in step S32 (step S34).

  For example, for each user group, the output product determination unit 19 selects a product group to which a trend tag has been added with a keyword tag that gives the sum of trend values, that is, the absolute number of products sold in a certain period, from 1 to 5. Output in descending order. The output product determination unit 19 sets the trend value for the product group to which the keyword tag with the sum total of trend values is the first, that is, the product group with the highest sales value, that is, for a certain period of time. The products with the absolute number of products sold to the top five are output as products, and the trend values are output in descending order. In addition, the product group to which the keyword tags having the second to fifth trend values are added are the products with the highest trend values, that is, the products that output the most sold products, that is, the products to be proposed.

  Thereby, the process which determines as a goods which proposes the goods which match a user's hobby preference is complete | finished. The control unit 10 acquires a product image or product information determined as a product to be output from the output product determination unit 19 and inputs it to the input / output unit 11. The input / output unit 11 outputs this result to an external display means or the like.

  The output result is displayed as shown in FIG. This is a display example for general users displayed on the top page of an EC (Electronic Commerce) site. Since this user belongs to the feminine user group, only the result of the feminine user group is displayed. For products with the keyword “Race” with the highest number of sales, product images with the top 5 sales are displayed. In addition, a product image and a description of the product are displayed for the product with the highest sales number among the products with the keyword tag “race”. For other keyword tags, the product image with the highest trend value is displayed. The description of the product can be displayed when the control unit 10 acquires the description of the product from the product DB 13 and outputs it from the input / output unit 11.

  According to the present embodiment, it is possible to propose a product that matches the user's hobbies and that is in line with the trend. Therefore, it is possible to propose a product that will motivate the user to purchase.

  In the present embodiment, based on the information about the user acquired from the user terminal and the purchase history, determination and display of the product to be output only for the user group corresponding to the user is performed. As shown in the embodiment, it is possible to determine the products to be output for all user groups and display only the results for the user groups corresponding to the users who have acquired information.

  A display example in this case is shown in FIG. As shown in FIG. 12, tabs of user groups of feminine, cool, sexy, casual, and natural are displayed. The tab is linked to a page of each user group, and the display content can be changed by the user selecting the tab. If it is determined that the user belongs to the feminine user group, the result for the feminine user group is displayed as a default. Thereby, since the user can know not only the result of the user group to which the user belongs, but also the result of the other user group, the trend can be easily grasped.

  In the present embodiment, in step S34, for the product group to which the keyword tag with the trend value sum being No. 1 is attached, the product with the trend value of No. 1 to No. 5 is output, and the trend The product group with the 2nd to 5th keyword tags having the sum total of values is determined as the product that outputs the product with the highest trend value, but the method for determining the product to be output is not limited to this. Based on the purchase history acquired in step S31, the user may extract unpurchased products from the products that have been ranked as shown in step S34, and output the extracted products. . Thereby, the user's willingness to purchase can be further evoked.

  The present invention may be provided as a system consisting of two devices, a trend analysis device or product recommendation device and a user terminal, or as a trend analysis device or product recommendation device including a user terminal function. Also good. Further, the present invention is not limited to the device, and can be provided as a program applied to the device.

It is the schematic of the trend analysis apparatus 1 of the 1st Embodiment of this invention. It is a flowchart which shows the flow of the registration process to DB of the trend analyzer 1. FIG. It is a figure explaining calculation of a feature-value. 4 is a flowchart showing a flow of trend analysis processing of the trend analysis device 1. It is a display example based on the result output from the trend analyzer 1. FIG. It is the schematic of the product recommendation apparatus 2 of the 2nd Embodiment of this invention. It is a flowchart which shows the flow of the merchandise proposal process of the merchandise recommendation apparatus. It is a display example based on the result output from the product recommendation device 2. It is the schematic of the goods recommendation apparatus 3 of the 3rd Embodiment of this invention. It is a flowchart which shows the flow of the product proposal process of the product recommendation apparatus 3. FIG. It is a display example based on the result output from the product recommendation device 3. It is a display example based on the result output from the product recommendation device 3. It is a figure explaining the conventional problem. It is a figure explaining the conventional problem. It is a figure explaining the conventional problem.

Explanation of symbols

1: Trend analysis device 2, 3: Product recommendation device 10: Control unit 11: Input / output unit 12: Memory control unit 13: Product DB 13: DB management unit 15: Image management unit 16: Decoder Encoder, 17: Feature amount extraction unit, 18: Trend value calculation unit, 19: Output product determination unit, 20: User DB, 21: DB management unit, 22: User group extraction unit, 31: Communication I / F

Claims (10)

Acquisition means for acquiring an image of the product and sales information of the product;
Means for analyzing the image of the product acquired by the acquisition means and calculating a feature amount of the product and / or creating a keyword tag indicating the feature of the product;
The keyword tag created by the sales information acquired by the acquisition means and the feature quantity calculated by the means for calculating the feature quantity from the product image corresponding to the sales information and / or the means for creating the keyword tag A product database that stores
Trend value calculation means for calculating a trend value for each feature amount and / or keyword tag, based on the sum of sales information of all products for which the same feature amount is calculated and / or the same keyword tag is created A trend value calculating means for calculating a value as a trend value;
Ranking means for ranking the feature quantities and / or keyword tags based on the trend value calculated by the trend value calculation means;
Output means for outputting the result of ranking performed by the ranking means;
A product analysis apparatus comprising: Based on specific information for which trend analysis is desired, a feature amount included in the specific information is extracted as a feature amount group from the product database, and / or a keyword tag included in the specific information is extracted from the product database. First extraction means for extracting as a keyword tag group,
The trend value calculation unit calculates a feature amount and / or a trend value of a keyword tag included in the feature amount group and / or the keyword tag group extracted by the first extraction unit. Product analysis device of description. The trend value calculation means calculates a trend value for each product based on the sales information,
The ranking means ranks the products for each feature quantity and / or keyword tag for which ranking has been performed based on the trend value for each product calculated by the trend value calculation means. The product analysis apparatus according to claim 1, wherein the product analysis apparatus is a product analysis apparatus. The means for calculating the feature amount extracts a color included in the image of the product acquired by the acquisition unit, calculates the most included color among the extracted colors as the feature amount,
The product analysis apparatus according to any one of claims 1 to 3, wherein the keyword tag creating means creates, as the keyword tag, text representing the most contained color.   The means for creating the keyword tag extracts at least one of the shape and pattern of the product by pattern recognition from the product image acquired by the acquisition unit, and extracts at least one of the extracted shape and pattern of the product. The product analysis apparatus according to claim 1, wherein a text to be expressed is created as the keyword tag. The product analysis apparatus according to any one of claims 1 to 5,
The product purchased by the user is stored in association with the purchase history including the feature amount calculated by the means for calculating the feature amount from the image of the product and / or the keyword tag created by the means for creating the keyword tag. User database to
User group extraction means for extracting users having similar hobbies and preferences as a user group based on the purchase history stored in the user database;
A feature amount included in the purchase history of the user group extracted by the user group extraction unit from the product database is extracted as a feature amount group, and / or a user group extracted by the user group extraction unit from the product database. A keyword tag included in the purchase history is extracted as a keyword tag group and a feature amount included in the purchase history of the user group is calculated from the product database and / or a keyword tag included in the purchase history of the user group A second extraction means for extracting the product for which
With
The trend value calculation means calculates a trend value of a feature quantity and / or keyword tag included in the feature quantity group and / or keyword tag group extracted by the second extraction means, and extracts by the second extraction means. Calculate the trend value of the selected product,
The ranking means ranks feature quantities and / or keyword tags included in the feature quantity group and / or keyword tag group extracted by the second extraction means, and is extracted by the second extraction means. Product ranking is performed for each feature quantity and / or keyword tag for which ranking is performed,
The output means ranks the feature quantities and / or keyword tags of a predetermined higher rank among the feature quantities and / or keyword tags ranked by the ranking means by the ranking means. A product proposal device that acquires products in a predetermined order from among the products, and outputs the acquired products as products to be proposed to a user. User purchase history acquisition means for acquiring a purchase history of a predetermined user;
Based on the purchase history acquired by the user purchase history acquisition means and the purchase history stored in the user database, the user group to which the predetermined user belongs from among the user groups extracted by the user group extraction means A user group specifying means for specifying
The feature quantity group and / or keyword tag group of the user group designated by the user group designation means is extracted from the feature quantity group and / or keyword tag group extracted by the second extraction means, and the designated A third extraction means for extracting the product for which the feature amount included in the feature amount group of the user group is calculated and / or the product for which the keyword tag included in the keyword tag group of the specified user group is created; Prepared,
The trend value calculating means calculates a feature value and / or a keyword tag trend value included in the feature quantity group and / or keyword tag group extracted by the third extracting means, and is extracted by the third extracting means. Calculate the trend value of the selected product,
The ranking means ranks feature quantities and / or keyword tags included in the feature quantity group and / or keyword tag group extracted by the third extraction means, and is extracted by the third extraction means. 7. The product proposal device according to claim 6, wherein product ranking is performed for each feature amount and / or keyword tag for which ranking is performed. A product having a higher predetermined rank among the products associated with the upper predetermined number of feature quantities and / or keyword tags among the feature quantities and / or keyword tags ranked by the ranking means. Unpurchased product extraction for extracting products not included in the purchase history of the predetermined user from the acquired products based on the purchase history of the predetermined user acquired from the database and acquired by the user purchase history acquisition means With means,
8. The product proposal apparatus according to claim 7, wherein the output unit outputs the product extracted by the unpurchased product extraction unit as a product to be proposed to a user. Obtaining a product image and sales information for the product;
Analyzing the acquired image of the product, calculating a feature amount of the product and / or creating a keyword tag indicating the feature of the product;
Associating the acquired sales information with the feature amount calculated by the step of calculating the feature amount from the image of the product corresponding to the sales information and / or the keyword tag created by the step of creating the keyword tag Saving to the product database,
A step of calculating a trend value for each feature amount and / or keyword tag, wherein the trend value is based on a total of sales information of all products for which the same feature amount is calculated and / or the same keyword tag is created Calculating as a value;
Ranking the feature quantities and / or keyword tags based on the calculated trend values;
Outputting the ranking results;
A product analysis method comprising:   A program for causing a computing device to execute the product analysis method according to claim 9.

Images