JP6031210B1 - Sales prediction device, sales prediction method, and program - Google Patents

Abstract

Conventionally, it has been impossible to predict sales of a product using an image such as a photograph of the product. A learning information storage unit that stores at least one learning information including learning image information that is information about a learning image that is an image of a product and sales result information about the sales result of the product; An image receiving unit that receives an image, analyzes the image, applies the analysis result to one or more pieces of learning information, and stores one or more pieces of sales record information included in one or more pieces of learning information. By using a sales prediction device that includes a sales prediction unit that is used and an output unit that outputs sales prediction information, the sales prediction of the product can be performed using an image such as a photograph of the product. [Selection] Figure 1

Description

  The present invention relates to a sales prediction device that performs sales prediction of products using images.

  Conventionally, when there are a plurality of models, and each consumer often selects and purchases a model according to his / her preference, there has been a sales prediction device that accurately predicts the number of sales of the target model (for example, Patent Document 1). In such a sales forecasting device, the characteristics of all consumers are set, one of the consumers set with the characteristics is extracted, the consumer's preference is determined, and if the consumer's preference is price-oriented A price-oriented model is executed. If the consumer's preference is model-oriented, the model-oriented model is executed. If the consumer's preference is specifications-oriented, a specification-oriented model is executed. Then, in such a sales prediction device, when the simulation for all consumers for which characteristics have been set is completed, a process of counting the number of sales for each model is performed.

  Conventionally, there has been a sales prediction device that can accurately perform sales prediction even with discontinuous sales performance information and short-term sales performance information less than one year (see, for example, Patent Document 2). Such a sales prediction device includes a sales performance information acquisition means for acquiring sales performance information related to the sales performance of each product, and a sales prediction means for predicting the sales volume of each product from the sales performance information acquired by the sales performance information acquisition means. In the sales forecasting device, the sales forecasting means aggregates the sales volume of each product for each day of the week from the sales performance information for a predetermined number of days, and calculates the sales volume of each product for each day of the week from the aggregated value. A sales prediction device characterized by prediction.

JP 2008-299786 A JP 2006-178777 A

  However, in the conventional sales prediction device, it has been impossible to make a sales prediction of the product using an image such as a photograph of the product. Conventionally, there has been no idea of inputting images and predicting sales of products.

  The sales prediction device according to the first aspect of the present invention is a learning information storage unit in which one or more pieces of learning information including learning image information that is information related to a learning image that is an image of a product and sales performance information that relates to the sales performance of the product are stored And an image receiving unit that receives an image of a product to be sold, an image that is received by the image receiving unit, the analysis result is applied to one or more learning information, and the sale of the product related to the image that is received by the image receiving unit The sales prediction device includes a sales prediction unit that acquires sales prediction information related to prediction using one or more sales performance information included in one or more learning information, and an output unit that outputs the sales prediction information.

  With this configuration, it is possible to predict sales of the product using an image such as a photograph of the product.

  In addition, the sales prediction device of the second invention further includes a correspondence information storage unit that stores correspondence information for associating the first sales prediction information and the second sales prediction information with respect to the first invention. The sales prediction unit analyzes the image received by the image reception unit, applies the analysis result to one or more learning information, and is the same type of information as the one or more sales performance information included in the one or more learning information. A first means for obtaining one sales forecast information and a second means for obtaining second sales forecast information using correspondence information from the first sales forecast information, and the output unit outputs the second sales forecast information It is a sales prediction device.

  With such a configuration, it is possible to output sales prediction information, which is sales prediction information of the product, which is different from the learned sales performance, using an image such as a photograph of the product.

  Further, in the sales prediction device according to the third aspect of the invention, the learning information is information learned by using deep learning from learning source information having a learning image and sales performance information. It is a sales prediction device.

  With this configuration, it is possible to predict the sales of the product with high accuracy using an image such as a photograph of the product.

  In the sales prediction device according to the fourth aspect of the present invention, in contrast to any one of the first to third aspects, the sales image and the image received by the image receiving unit are sales predictions that are photographs of one or more products. Device.

  With this configuration, it is possible to predict the sales of the product using a photograph of the product.

  Further, in the sales prediction device of the fifth invention, the image received by the image receiving unit includes images of two or more products for any one of the first to fourth inventions, Two or more partial images that are images of two or more products are acquired from the image received by the image receiving unit, each of the two or more partial images is analyzed, and the analysis result is converted into one or more learning information for each partial image. Apply and obtain the sales prediction information of the product related to the image received by the image receiving unit using one or more sales performance information included in one or more learning information, the output unit sales prediction of two or more of each product It is a sales prediction device that outputs information.

  With this configuration, sales prediction of two or more products included in an image can be performed together.

  The sales prediction device according to the sixth aspect of the present invention provides the learning information storage unit according to any one of the first to fifth aspects, wherein the learning information storage unit is a learning value that is two or more attribute values of the learning image acquired from the learning image. One or more pieces of learning information including image information and sales performance information are stored, and the sales prediction unit analyzes the image received by the image receiving unit, acquires an analysis result that is two or more attribute values, and stores the analysis result. A sales prediction device that applies to one or more pieces of learning information and acquires sales prediction information related to the sales prediction of a product related to an image received by the image receiving unit using one or more pieces of sales record information included in the one or more pieces of learning information. is there.

  With this configuration, it is possible to predict sales of the product using an image such as a photograph of the product.

  According to the sales prediction device of the present invention, it is possible to make a sales prediction of a product using an image such as a photograph of the product.

Block diagram of sales forecasting apparatus A in the first embodiment Block diagram of learning device B A flowchart for explaining the operation of the sales prediction apparatus A A flowchart for explaining the operation of the learning device B Figure showing the same image / sales performance management table Figure showing the correspondence information management table The figure which shows the example of the image of the same sales forecast object Figure showing the same output example Figure showing the same output example The figure which shows the example of the image of the same sales forecast object Figure showing the same output example The figure which shows the information of the pair of the same image information and sales performance information The figure which shows the example of the merchandise information which the sales prediction apparatus A receives Figure showing the same output example Overview of the computer system Block diagram of the computer system

  Hereinafter, embodiments of a sales prediction device and the like will be described with reference to the drawings. In addition, since the component which attached | subjected the same code | symbol in embodiment performs the same operation | movement, description may be abbreviate | omitted again.

  (Embodiment 1)

  In the present embodiment, the received image is applied to learning information learned by using two or more sets of images and sales record information, and sales prediction information corresponding to the images is acquired and output. The prediction device will be described. In this specification, the sales performance is preferably the sales performance of products or services, but the sales performance of products or services, the performance of third-party actions on images such as photographs (for example, SNS) It may be considered that it includes the results of third-party reactions to posted photos. The sales forecast information is preferably information related to the sales performance of the product or service, but information related to the rental performance of the product or service, or information related to the performance of a third party action on an image such as a photograph. And so on.

  Further, in the present embodiment, a sales prediction apparatus will be described in which sales performance information to be learned and sales prediction information to be output are different types of information. Further, in this specification, the sales forecast is preferably a forecast related to the sale of goods or services, but a forecast related to the rental of goods or services, a forecast related to third party actions on images such as photographs, etc. You can think of it as including.

  In the present embodiment, a sales prediction apparatus that outputs sales prediction information for two or more products in an image will be described. Note that the products here include so-called services and the entire sales target.

  Furthermore, in the present embodiment, a learning device that learns learning information used by the sales prediction device A will be described.

  FIG. 1 is a block diagram of a sales prediction apparatus A in the present embodiment. The sales prediction device A includes a storage unit 1, a reception unit 2, a processing unit 3, and an output unit 4. The storage unit 1 includes a learning information storage unit 11 and a correspondence information storage unit 12. The reception unit 2 includes an image reception unit 21. The processing unit 3 includes a sales prediction unit 31. The sales prediction unit 31 includes a first unit 311 and a second unit 312.

  The learning information storage unit 11 constituting the storage unit 1 stores one or more learning information. The learning information includes learning image information and sales performance information. It is preferable that the one or more pieces of learning information stored in the learning information storage unit 11 are information accumulated by the learning device B described later.

  The learning image information is information related to one or more learning images that are images of products. The learning image is usually an image obtained by photographing a product, but it may be an image created by using an editor or the like for the design of the product. That is, the data structure of the learning image does not matter. The learning image information is, for example, one or more attribute values of the learning image acquired from the learning image. The learning image information is, for example, one or two or more attribute values acquired from the learning image obtained by image analysis of the learning image. Here, the attribute value is, for example, a statistical processing amount of color, a pixel value (for example, an average value, a median value, etc.), a statistical processing amount of a pixel value of a background color, and a statistical processing of a pixel value of a product image in the image Included in the learning image, feature quantity such as information on the difference between the quantity, the statistical processing quantity of the pixel value of the product image in the image and the statistical processing quantity of the pixel value of the partial image excluding the product in the image One or more type identifiers for identifying the type of product. The learning image information may include, for example, text, information that can be acquired from text, vector data, or the like. The text is, for example, an attribute value of the product (for example, price, size, etc.), a description of the product, and the like. The learning image information may be, for example, one or more learning images themselves.

  The sales performance information is information related to the sales performance of the product. Sales performance information includes, for example, the total number of sales, the number of units sold, the number of people who purchased the product, whether the product was sold, whether it was made, whether it was profitable, the sales period, the sales area, the attributes of the purchaser Information (for example, information on gender ratio, age group), price, product identifier for identifying a product to be purchased in combination. The sales performance information may be of any type or content as long as it is information related to the sales performance of the product. Further, the sales performance information may be information that does not lead to sales. The information that does not lead to sales is, for example, information related to the attention level of the product. The information related to the attention level of the product is, for example, the number of browsing of product information such as web pages and images related to the product, the number of people who have viewed product information such as web pages and images related to the products, and the like.

  It is preferable that the one or more pieces of learning information have a multilayered neural network structure. In addition, the one or more pieces of learning information are preferably information learned by deep learning from learning source information having a learning image and sales performance information, for example. The learning information is information including sales results of products. When the one or more learning information is the structure of a neural network having a multilayer structure, for example, the input to the neural network is one or two or more images, and one or more kinds of sales performance information described above. A neural network having a multilayer structure is obtained by inputting one or two or more images and one or more kinds of sales record information described above as input signals to a deep learning node and learning them.

  The one or more pieces of learning information may be information having one or more features (attribute values) acquired from the image and sales performance information. Such one or more pieces of learning information are used by machine learning such as SVR, for example.

  The correspondence information storage unit 12 stores one or more correspondence information. The correspondence information is information for taking correspondence between the first sales forecast information and the second sales forecast information. Both the first sales prediction information and the second sales prediction information are information related to sales prediction, but are different types of information. The correspondence information includes, for example, a product type identifier and a condition for determining that the product can be sold. The condition for determining that the item can be sold is, for example, “determined that the item can be sold if the first sales prediction information is greater than or equal to a threshold value or greater than the threshold value”. The correspondence information includes, for example, a product type identifier and a condition for determining that the product cannot be sold. The condition for determining that the item cannot be sold is, for example, “determining that the item cannot be sold when the first sales prediction information is equal to or smaller than the threshold value or smaller than the threshold value”. It is preferable that the above condition exists for each product type identifier. Further, it is preferable that the threshold value included in the condition varies depending on the product type identifier. The correspondence information is, for example, an arithmetic expression for calculating the second sales forecast information using the first sales forecast information as a parameter.

  The accepting unit 2 accepts instructions and information. The instructions, information, and the like are, for example, one or more images, an operation start instruction, and the like. Here, reception means reception of information input from an input device such as a keyboard, mouse, touch panel, reception of information transmitted via a wired or wireless communication line, recording on an optical disk, magnetic disk, semiconductor memory, or the like. It is a concept including reception of information read from a medium. Also, the input means such as instructions and information may be anything such as a numeric keypad, keyboard, mouse or menu screen. The accepting unit 2 can be realized by a device driver for input means such as a numeric keypad and a keyboard, control software for a menu screen, and the like.

  The image receiving unit 21 constituting the receiving unit 2 receives one or more images. The image received here is an image of a product for which sales are predicted. Further, the image received here is not an image corresponding to the product image information. In addition, it is suitable for the image reception part 21 to receive the photograph image | photographed with the said camera including a camera. The photograph is usually a photograph of a product for which sales forecast information is desired.

  The processing unit 3 performs various processes. The various types of processing are, for example, sales prediction processing performed by the sales prediction unit 31. The various processes are, for example, learning processes to be described later. When the processing unit 3 performs the learning process, the processing unit 3 includes a learning unit 61 and a storage unit 62 described later. Further, the processing unit 3 may acquire one or more images (photographs or the like) corresponding to sales that are so large that the sales performance information satisfies a predetermined condition from the storage unit 1. For example, the processing unit 3 may acquire from the storage unit 1 one or more images (photos, etc.) corresponding to products for which the sales total information included in the sales performance information is greater than or equal to the threshold.

  The sales prediction unit 31 analyzes one or more images received by the image receiving unit 21, applies the acquired analysis result to one or more learning information, and relates to sales prediction of a product related to the image received by the image receiving unit 21. Sales forecast information is acquired using one or more sales performance information which one or more learning information has. The analysis result is, for example, one or more attribute values of the image. The sales prediction unit 31 acquires the sales prediction information by, for example, deep learning. The sales prediction unit 31 may acquire the sales prediction information by machine learning other than deep learning, such as SVR and decision tree. Note that the sales prediction unit 31 analyzes, for example, the image received by the image reception unit 21, acquires one or more attribute values of the image, and sets the one or more attribute values to one or more of the learning information storage unit 11. Applying to learning information, information of the same type as sales performance information possessed by the learning information is acquired by an algorithm such as deep learning or SVR. Note that the same type of information as the sales performance information is sales prediction information.

  The sales prediction unit 31 applies information including one or more images received by the image receiving unit 21 to one or more learning information in the learning information storage unit 12, and uses one or more sales performance information by an algorithm such as deep learning. You may get

  In addition, it is suitable for the sales prediction part 31 to acquire sales prediction information, using only an image as an input. That is, it is preferable that the sales prediction unit 31 applies only the image to one or more pieces of learning information in the learning information storage unit 11 and acquires the sales prediction information.

  When the image received by the image receiving unit 21 includes two or more merchandise symbols, the sales prediction unit 31 operates as follows, for example. That is, the sales prediction unit 31 acquires two or more partial images that are images of two or more products from the image received by the image reception unit 21, analyzes the two or more partial images, and for each partial image, The analysis result is applied to one or more learning information, and the sales prediction information of the product related to the image received by the image receiving unit 21 is acquired using one or more sales performance information included in the one or more learning information. Note that the partial image is a part of an image (for example, a photograph) and is a portion including (photographed) one product. In addition, the case where the image includes a pattern of two or more products is a case where the image is a photograph of two or more products taken.

  Further, the sales prediction unit 31 analyzes one or more images received by the image receiving unit 21, acquires an analysis result that is one or more attribute values, and uses the analysis result as one or more learning information (for example, The sales forecast information related to the sales forecast of the product related to the image received by the image receiving unit 21 using one or more sales record information included in one or more learning information. Is preferred. Applying the analysis result to one or more pieces of learning information is, for example, inputting the analysis result to a multi-layer neural network.

  Note that the sales prediction unit 31 may be configured by the first means 311 and the second means 312. In such a case, the first means 311 analyzes one or more images received by the image receiving unit 21, applies the analysis result to one or more learning information, and one or more sales results possessed by the one or more learning information. First sales forecast information that is the same type of information is acquired. Moreover, the 2nd means 312 acquires 2nd sales prediction information using correspondence information from 1st sales prediction information.

  The output unit 4 outputs the sales prediction information acquired by the sales prediction unit 31. The output unit 4 may output the second sales forecast information acquired by the second unit 312.

  Moreover, the output part 4 may output the sales prediction information of each two or more goods. In such a case, it is preferable that the output unit 4 outputs the sales prediction information in association with the type identifier indicating the type of each of two or more products.

  The output unit 4 may output one or more images (photographs, etc.) corresponding to sales that are so large that the sales performance information satisfies a predetermined condition. Further, the output unit 4 may output, for example, one or more images (photographs, etc.) corresponding to products for which the total sales number included in the sales performance information is greater than or equal to the threshold.

  Here, output refers to display on a display, projection using a projector, printing with a printer, sound output, transmission to an external device, storage in a recording medium, other processing devices or other programs, etc. This is a concept that includes delivery of processing results to

  FIG. 2 is a block diagram of learning device B in the present embodiment. The learning device B includes a storage unit 5 and a processing unit 6. The storage unit 5 includes a learning information storage unit 12, an image storage unit 51, and a sales performance information storage unit 52. The processing unit 6 includes a learning unit 61 and a storage unit 62.

  The storage unit 5 constituting the learning device B can store various types of information. Various types of information include images, sales performance information, and the like.

  The image storage unit 51 stores one or more images. The image includes an image of one or more products. The image is, for example, a photograph showing one or more products.

  The sales record information storage unit 52 stores sales record information in association with each image in the image storage unit 51.

  The processing unit 6 performs various processes. The various processes are processes performed by the learning unit 61, the storage unit 62, and the like, for example.

  The learning unit 61 performs a learning process using one or more images in the image storage unit 51 and one or more sales result information in the sales result information storage unit 52, and acquires one or more learning information. The learning unit 61 performs, for example, a learning process using a deep learning algorithm using one or more images in the image storage unit 51 and one or more sales record information in the sales record information storage unit 52. Get information. Since deep learning is a known technique, a detailed description thereof will be omitted.

  For example, the learning unit 61 analyzes each image in the image storage unit 51 and acquires one or more attribute values from each image. Then, the learning unit 61 configures a neural network having a multilayer structure that receives one or more attribute values and outputs sales performance information. In this case, the neural network having a multilayer structure is one or more pieces of learning information. The one or more attribute values acquired from the image include, for example, a type identifier, a color, a statistical processing amount of the pixel value (for example, an average value, a median value, etc.), a statistical processing amount of the pixel value of the background color, Information on the difference between the pixel value statistical processing amount of the product image, information on the difference between the pixel processing statistical value of the product image in the image and the pixel processing statistical value of the partial image excluding the product in the image, etc. And the like. Note that the deep learning algorithm in which the learning unit 61 forms a multilayered neural network is a known technique, and thus detailed description thereof is omitted.

  For example, the learning unit 61 analyzes each image in the image storage unit 51 and acquires one or more attribute values from each image. Then, the learning unit 61 acquires learning information having one or more attribute values (which may be referred to as one or more features) and sales performance information corresponding to the images for each image.

  The accumulation unit 62 accumulates one or more pieces of learning information acquired by the learning unit 61 in the storage unit 5. The one or more learning information accumulated here is one or more learning information stored in the learning information storage unit 11. The accumulating unit 62 accumulates, for example, the multilayered neural network acquired by the learning unit 61 in the storage unit 5.

  The storage unit 1, the learning information storage unit 11, the correspondence information storage unit 12, the storage unit 5, the learning information storage unit 11, and the image storage unit 51 are preferably non-volatile recording media. It is feasible. The process in which information is stored in the storage unit 1 or the like is not limited. For example, information may be stored in the storage unit 1 or the like via a recording medium, information transmitted via a communication line or the like may be stored in the storage unit 1 or the like, Alternatively, information input via the input device may be stored in the storage unit 1 or the like.

  The processing unit 3, the sales prediction unit 31, the first unit 311, the second unit 312, the processing unit 6, the learning unit 61, and the storage unit 62 can be usually realized by an MPU, a memory, or the like. The processing procedure of the processing unit 3 or the like is usually realized by software, and the software is recorded on a recording medium such as a ROM. However, it may be realized by hardware (dedicated circuit).

  The output unit 4 may be considered as including or not including an output device such as a display or a speaker. The output unit 4 can be realized by driver software for an output device or driver software for an output device and an output device.

  Next, operation | movement of the sales prediction apparatus A is demonstrated using the flowchart of FIG. In the flowchart of FIG. 3, it is assumed that the sales prediction apparatus A performs sales prediction processing for one image. Then, when the sales prediction apparatus A receives a plurality of images, the operation of the following flowchart may be repeated for the number of images.

  (Step S301) The image receiving unit 21 determines whether one or more images have been received. If one or more images are received, the process goes to step S302. If no image is received, the process returns to step S301. Note that the image receiving unit 21 may receive one or more images by the image receiving unit 21 taking a picture. That is, when a photograph is taken, the sales prediction device A may immediately perform the processing of the flowchart of FIG. 3 and output the sales prediction information for the product shown in the photograph.

  (Step S302) The sales prediction unit 31 performs an analysis process on the image received in step S301. The sales prediction unit 31 usually analyzes the image received in step S301 and acquires one or more attribute values. The one or more attribute values usually include a type identifier (for example, jacket, pants, T-shirt, etc.) that identifies the type of product included in the image (for example, shown in the photograph). Moreover, when the image includes partial images of two or more products, the sales prediction unit 31 usually acquires one or more attribute values for each partial image.

  (Step S303) The sales prediction unit 31 substitutes 1 for a counter i.

  (Step S304) The sales prediction unit 31 determines whether or not the image of the i-th product is included in the image received in step S301. When the image of the i-th product is included, the process goes to step S305, and when the image of the i-th product is not included, the process returns to step S301. Whether or not the image of the i-th product is included in the image is determined, for example, by whether or not the i-th type identifier is present in one or more attribute values acquired in step S302. Is the same as

  (Step S305) The sales prediction unit 31 applies the analysis result of the partial image of the i-th product to one or more pieces of learning information in the learning information storage unit 11, and acquires first sales prediction information. When there is only one product in the image, the partial image of the i-th product is usually the entire image. However, even when the number of products in the image is one, the sales prediction unit 31 may cut out the product region from the image and perform image analysis processing on the image in the cut-out region. In addition, when there is one product in the image, the analysis result of the partial image of the i-th product is the analysis result of the entire image.

  (Step S306) The sales prediction unit 31 determines whether or not to acquire the second sales prediction information by applying the correspondence information to the first sales prediction information acquired in step S305. If the second sales forecast information is acquired, the process goes to step S307. If the second sales forecast information is not acquired, the process goes to step S308. Whether to apply the correspondence information may be determined in advance, may be determined based on whether the correspondence information exists in the correspondence information storage unit 12, or the type information of the i th product. Alternatively, it may be determined based on whether or not correspondence information corresponding to the product type information in the image exists in the correspondence information storage unit 12.

  (Step S307) Sales prediction part 31 applies correspondence information to the 1st sales prediction information, and acquires the 2nd sales prediction information.

  (Step S308) The output unit 4 outputs sales prediction information. The sales prediction information output here is the second sales prediction information acquired in step S307 or the first sales prediction information acquired in step S305. If the second sales prediction information is not acquired in step S307, the output unit 4 outputs the first sales prediction information acquired in step S305. When the second sales prediction information is acquired in step S307, the output unit 4 outputs the second sales prediction information acquired in step S307. When the second sales prediction information is acquired in step S307, the output unit 4 may also output the first sales prediction information acquired in step S305. Further, it is preferable that the output unit 4 outputs the sales prediction information in association with the product type identifier in the image.

  (Step S309) The sales prediction unit 31 increments the counter i by 1. The process returns to step S304.

  In the flowchart of FIG. 3, the process ends when the power is turned off or the process is terminated.

  In the flowchart of FIG. 3, the receiving unit 2 may receive a plurality of photographs for the same product. In such a case, the processing unit 3 may determine that the photo to which the best sales prediction information is output is the best photo. And it is suitable for the output part 4 to output this judgment result.

  Next, operation | movement of the learning apparatus B is demonstrated using the flowchart of FIG.

  (Step S401) The learning unit 61 substitutes 1 for a counter i.

  (Step S402) The learning unit 61 determines whether one or more images of the i-th product exist in the image storage unit 51. When the i-th image exists, the process goes to step S403, and when it does not exist, the process goes to step S406. Here, the learning unit 61 may determine whether or not the i-th sales performance information exists in the sales performance information storage unit 52. And when the i-th sales performance information exists, it goes to step S403, and when it does not exist, you may go to step S406.

  (Step S403) The learning unit 61 performs an image analysis process on one or more images of the i-th product and acquires an analysis result. Then, the learning unit 61 usually obtains an analysis result that is one or more attribute values of the i-th image. Note that processing for obtaining one or more attribute values from an image can be performed by a known technique.

  (Step S404) The accumulating unit 62 accumulates the analysis result acquired in step S403 in the storage unit 5.

  (Step S405) The learning unit 61 increments the counter i by 1. The process returns to step S402.

  (Step S406) The learning unit 61 configures learning information using the analysis result accumulated in step S404 and the sales result information of each image. Note that the learning unit 61 preferably uses a deep learning algorithm to construct a neural network having a multilayer structure that outputs the sales result information of each image using the analysis result accumulated in step S404. When one or more images are input to such a multi-layered neural network, sales forecast information is output. The sales forecast information is usually the same type of information as the sales performance information. A neural network having a multilayer structure is an example of learning information.

  (Step S407) The storage unit 62 stores the learning information configured in step S406 in the learning information storage unit 11. The process ends. The learning information configured in step S406 is, for example, a multilayered neural network.

  Hereinafter, specific operations of the sales prediction device A and the learning device B in the present embodiment will be described.

  Now, it is assumed that the storage unit 5 of the learning device B stores the image / sales record management table shown in FIG. The image / sales record management table is a table that stores information on sales results of images and products in the image. The image / sales result management table includes “ID”, “image”, “type information”, and “sales result information”. “Sales result information” includes “total number of sales”, “number of sales in April”, “PV”, “PVUU”, and “CVUU”. “ID” is information for identifying a record. “Type information” is information for identifying the type of product. “PV” is information indicating the degree of attention of a product, and here is the number of browsing web pages of the product. “PVUU” is the number of viewers of the product web page. “CVUU” is the number of people who have purchased the product.

  Further, it is assumed that the correspondence information storage unit 12 of the sales prediction apparatus A stores the correspondence information management table shown in FIG.

In this situation, the following four specific examples will be described. In specific example 1, a specific operation of learning device B will be described. In the second specific example, the specific operation of the sales prediction apparatus A when one product is shown in the input image will be described. In specific example 3, a specific operation of sales predicting apparatus A when two or more products are shown in the input image will be described. In specific example 4, the operation of learning device B when learning device B learns using information including information other than product images and acquires one or more pieces of learning information will be described. Moreover, in the specific example 4, the process which the sales prediction apparatus A performs using the 1 or more learning information which the said learning apparatus B learned is demonstrated.
(Specific example 1)

  First, it is assumed that the user inputs an instruction to start learning to the learning device B. Then, the learning device B receives an instruction to start such learning. Then, the learning unit 61 performs image analysis processing on one or more images included in each record of FIG. 5, and obtains an analysis result that is one or more attribute values for each image of each record of FIG. To do. Then, the learning device B forms a neural network having a multilayer structure by deep learning using the analysis result, the type information of each record, and the sales performance information. Note that the type information of each record is preferably information acquired by image analysis processing.

Then, the storage unit 62 stores the multilayered neural network (one or more learning information) acquired by the learning unit 61 in the storage unit 5. The storage unit 5 also stores images used by the learning unit 61. It is assumed that the neural network and images having a multilayer structure stored in the storage unit 5 are moved or copied to the learning information storage unit 11 of the sales prediction apparatus A.
(Specific example 2)

  Assume that the user gives the image of FIG. The image in FIG. 7 is a photograph of a jacket that will be mass-produced in the future. And the image reception part 21 receives the image of FIG.

  Next, the sales prediction unit 31 performs an analysis process on the received image. Then, the sales prediction unit 31 acquires one or more attribute values. Here, it is assumed that the sales prediction unit 31 acquires attribute values such as type information “jacket”, color “blue”, shooting method or direction “oblique direction”.

  Next, the sales prediction unit 31 applies the analysis result (type information “jacket”, color “blue”, shooting method or direction “oblique direction”, etc.) to the neural network having the multilayer structure of the learning information storage unit 11. Suppose that sales forecast information “21” is acquired.

  Next, the sales prediction unit 31 applies the type information “jacket” and the sales prediction information “21” to the correspondence information management table of FIG. 6 to obtain the second sales prediction information “not sellable”. That is, the sales prediction unit 31 determines that the condition information “if (total number of sales <= 100)” indicated by the correspondence information of the record of “ID = 2” corresponding to the type information “jacket” matches the second sales. Get prediction information "Not sell".

  Next, the output unit 4 outputs “sales forecast information“ 21 ”” and “second sales forecast information“ not sellable ”, which are sales forecast information, in association with the type information“ jacket ”. An example of such output is shown in FIG. The output unit 4 indicates that “this“ $ type information ”is“ $ second sales forecast information ”.” And “this“ $ type information ”has a predicted total sales number“ $ sales forecast information ”. ”Is stored. “$...” In the template of the output sentence is a variable.

  Further, the processing unit 3 obtains an image corresponding to the type information “jacket” from the learning information storage unit 11, which is an image corresponding to a product that is so large that the total number of sales satisfies a predetermined condition. And the output part 4 outputs a screen as shown in FIG. 9 using this image. In addition, the output section 4 is “← Your“ $ type information ”photo.” “Sold“ $ type information ”is the following photo.” “Like the photo below, how to take a photo Suppose you have a template for the output sentence, such as "Is it better to be devised?" The output unit 4 also determines the total number of sales only when it is determined that the sales prediction information or the second sales prediction information acquired by the sales prediction unit 31 satisfies a predetermined condition (usually in the case of a determination that it cannot be sold). It is preferable to output an image of a product with a large number of items. Note that the attribute value of the shooting method or direction of the jacket image output in FIG. 9 is, for example, “front”. In FIG. 9, a comment may be output using “front” which is a shooting method or direction. Such a comment is, for example, “How about taking a picture in the front direction”?

  As described above, according to this specific example, it is possible to make a sales prediction by simply inputting an image such as a photograph of a product that is a product before sales or before mass production, and for which a product for which sales are to be predicted.

In addition, according to the present specific example, a photograph that serves as a reference for increasing the number of sales is presented only by inputting a photograph of a product before sale or the like. Note that the user who sees the output example of FIG. 9 thinks that it is better to photograph the jacket to be sold from the front in order to increase the number of jackets sold.
(Specific example 3)

  Assume that the user gives the image of FIG. Then, the image reception part 21 receives the image of FIG.

  Next, the sales prediction unit 31 performs the following analysis processing on the image of FIG. That is, the sales prediction unit 31 determines that two products are shown in the image of FIG. Next, the sales prediction unit 31 cuts out an image of the product for each of the two products in the image of FIG. 10 (1001 and 1002), and performs image analysis. And sales prediction part 31 acquires one or two or more attribute values for every goods. Here, it is assumed that the sales prediction unit 31 acquires attribute values such as type information “jacket” for each product.

  Next, the sales prediction unit 31 applies the analysis result (type information “jacket” or the like) to the neural network having the multilayer structure of the learning information storage unit 11 for each product, and acquires the sales prediction information.

  Next, the output unit 4 outputs sales prediction information for each product. An example of such output is shown in FIG.

  As described above, according to this specific example, it is possible to obtain sales predictions for a plurality of products at the same time by simply inputting a photograph of a plurality of products before sales.

As described above, according to the present embodiment, sales prediction of a product can be performed using an image such as a photograph of the product.
(Specific example 4)

  Now, in the storage unit 5 of the learning apparatus B, information of a pair of image information and sales performance information shown in FIG. 12 is stored. In FIG. 12, “pic” is a product image in the image storage unit 51. “Pv” is information indicating the degree of attention of a product, and here is the number of browsing web pages of the product. “Cv” is the number of products sold. “Pvuu” is the number of viewers of the product web page. “Cvuu” is the number of people who purchased the product. The “sales duration” is a period in which the sale of the product is continued. “Annual income” is the line of annual income of the customer targeted by the product or the annual income of the customer who purchased the product. “Educational background” is the educational background of the customer targeted by the product or the educational background of the customer who purchased the product. “Child” is whether the customer targeted by the product has a child or whether the customer who purchased the product has a child. The “fashion category” is a fashion category of a product, and may be called a product type.

  In such a situation, it is assumed that the user inputs an instruction to start learning to the learning device B. Then, the learning device B receives an instruction to start such learning. Then, the learning unit 61 learns the image information and sales performance information of FIG. 12 by a deep learning algorithm, acquires learning information, and accumulates it in the learning information storage unit 12.

  Next, it is assumed that the user inputs input information including images of three products shown in FIG.

  Then, the reception part 2 receives the record of three goods shown in FIG. Next, the sales prediction unit 31 applies the records of the three products received by the receiving unit 2 to one or more pieces of learning information in the learning information storage unit 12, and uses one or more pieces of sales record information by a deep learning algorithm. Get. And the output part 4 outputs the acquired 1 or more sales prediction information.

An example of such output is shown in FIG. In FIG. 14, the record and sales forecast information of FIG. 12 are output. In FIG. 14, one or more pieces of sales prediction information are indicated by probabilities.
(Experiment)

  An experiment for performing sales prediction using the sales prediction apparatus A in the present embodiment was performed. As a result, good results were obtained as shown below.

  When an image of the product type “Tops” is input to the sales prediction device A, the probability that the prediction (“PV” above) of “Tops” will become a popular product is won (predicted to become a popular product and popular When it became a product, it was predicted that it would not become a popular product, and when it became a popular product) was 81.60%. In addition, when the image of the product type “one piece” is input to the sales prediction device A, the probability that it is predicted whether or not it will be a popular product (if the product is predicted to become a popular product, 76.20%) when it was predicted that it would not be a product and it became a popular product. In addition, when an image of the product type “suit” is input to the sales prediction device A, the probability that it is predicted whether or not it will be a popular product (if the product is predicted to become a popular product, 88.20%) when it was predicted that it would not be a product and it became a popular product. In addition, when the image of the product type “outer” is input to the sales prediction device A, the probability that the product is predicted to be a popular product or not (the popularity is predicted when the product becomes a popular product when predicted to be a popular product) 80% was 40.40% when it was predicted that it would not be a product and it became a popular product. In addition, when the image of the product type “bottom” is input to the sales prediction device A, the probability that it is predicted whether or not it will be a popular product (if the product is predicted to become a popular product and becomes popular, 76.50%) when it was predicted that it would not be a product and it became a popular product. In addition, when an image of the product type “shoes / socks / shoe-related goods” is input to the sales prediction device A, the probability that the product is predicted to be a popular product (predicted to be a popular product and a popular product In the case of becoming a popular product in anticipation of not becoming a popular product), it was 65.00%. In addition, when an image of the product type “Cosmetics” is input to the sales prediction device A, the probability that it is predicted whether or not it will be a popular product (if the product is predicted to become a popular product and becomes popular, 100.00%) when it was predicted that it would not be a product and became a popular product.

In addition, when an image of the product type “Tops” is input to the sales prediction apparatus A, the probability that the prediction of whether or not “Tops” is a product to be purchased (the above-mentioned CV) has been won (the product to be purchased). If it was a product that was bought with a prediction, it was 74.20% if it was not a product that was bought with a prediction that it was not a product to be bought. In addition, when an image of the product type “one piece” is input to the sales prediction apparatus A, the probability of the prediction of whether or not “one piece” is a product to be purchased (the above-mentioned CV) has been won (the product to be bought). If it was a product that was purchased with a prediction that the product was not a product that was purchased with no prediction, it was 74.10%. In addition, when an image of the product type “suit” is input to the sales prediction apparatus A, the probability that the prediction of whether or not the “suit” is a product to be purchased (the above CV) has been won (the product to be purchased). If it was a product that was bought with a prediction, it was 63.20% if it was not a product that was bought with a prediction that it was not a product to be bought. In addition, when an image of the type of product “outer” is input to the sales prediction apparatus A, the probability that the prediction of whether or not “outer” is a product to be purchased (the above-mentioned CV) has been won (the product to be purchased). If it was a product that was bought with a prediction, it was 70.00% if it was not a product that was purchased with a prediction that it was not a product to be bought. In addition, when an image of the product type “bottom” is input to the sales prediction apparatus A, the probability that the prediction of whether or not “bottom” is a product to be purchased (the above CV) has been won (the product to be purchased). And 73.20% (in the case of a product that was predicted and not bought). In addition, when an image of “Shoes / Socks / Shoes Related Goods” is input to the sales prediction device A, whether or not “Shoes / Socks / Shoes Related Goods” is purchased (CV above) Is 70.60% (when it is a product that is bought with a prediction that it is a product to be bought, when it is not a product that is bought with a prediction that it is not a product to be bought). In addition, when an image of the product type “bag” is input to the sales prediction apparatus A, the probability that the prediction of whether or not the “bag” is a product to be purchased (the above CV) has been won (the product to be purchased). And 75.80% in the case of a product that was purchased with a prediction that it was not a product that was purchased without a prediction.
In addition, when an image of the type of product “Cosmetics” is input to the sales prediction device A, the probability that “Cosmetics” is a product to be bought (CV above) has been predicted (the product to be bought). If it was a product that was bought with a prediction, it was 86.70% if it was not a product that was bought without a prediction.

  Further, the processing in the present embodiment may be realized by software. Then, this software may be distributed by software download or the like. Further, this software may be recorded and distributed on a recording medium such as a CD-ROM. This also applies to other embodiments in this specification. In addition, the software which implement | achieves the sales prediction apparatus A in this Embodiment is the following programs. In other words, in this program, the computer-accessible recording medium stores one or more pieces of learning information including learning image information that is information related to a learning image that is an image of a product and sales result information that is related to the sales result of the product. A learning information storage unit, and the computer analyzes an image receiving unit that receives an image that is an image of a product to be sold and is not an image corresponding to the product image information, and an image received by the image receiving unit. The analysis result is applied to the one or more learning information, and the sales prediction information related to the sales prediction of the product related to the image received by the image receiving unit is used as one or more sales performance information included in the one or more learning information. It is a program for functioning as a sales prediction part acquired using, and an output part which outputs the sales prediction information.

  In the above program, the recording medium further includes a correspondence information storage unit that stores correspondence information for associating the first sales prediction information with the second sales prediction information, and the sales prediction unit includes the image reception unit. The image received by the department is analyzed, the analysis result is applied to the one or more learning information, and the first sales prediction information that is the same type as the one or more sales performance information included in the one or more learning information is acquired. And second means for obtaining second sales forecast information from the first sales forecast information using the correspondence information, and the output unit outputs the second sales forecast information. It is preferable that the program is a program that causes a computer to function.

  In the above program, the learning information is preferably a program that causes a computer to function as information learned using deep learning from learning source information having a learning image and sales performance information. .

  In the above program, it is preferable that the learning image and the image received by the image receiving unit are a program that causes a computer to function as a photograph of one or more products.

  In the above program, the image received by the image receiving unit includes images of two or more products, and the sales prediction unit is an image of each of the two or more products from the image received by the image receiving unit. Articles related to an image obtained by acquiring two or more partial images, analyzing each of the two or more partial images, applying the analysis result to the one or more learning information for each partial image, and receiving the image reception unit The sales forecast information is acquired using one or more sales record information included in the one or more learning information, and the output unit functions as a computer that outputs the sales forecast information of each of the two or more products. It is preferable that the program is to be executed.

  In the program, the learning information storage unit stores one or more pieces of learning information including learning image information and sales performance information that are two or more attribute values of the learning image acquired from the learning image, The sales prediction unit analyzes the image received by the image receiving unit, acquires an analysis result that is two or more attribute values, applies the analysis result to the one or more learning information, and the image receiving unit receives the analysis result. It is preferable that the program is a program that causes a computer to function as one that acquires sales prediction information related to sales prediction of a product related to an image using one or more sales performance information included in the one or more learning information.

  FIG. 15 shows the external appearance of a computer that executes the programs described in this specification and realizes the sales prediction apparatus and the like of the various embodiments described above. The above-described embodiments can be realized by computer hardware and a computer program executed thereon. FIG. 15 is an overview of the computer system 300, and FIG. 16 is a block diagram of the system 300.

  In FIG. 15, the computer system 300 includes a computer 301 including a CD-ROM drive 3012, a keyboard 302, a mouse 303, a monitor 304, and a camera 307.

  In FIG. 16, a computer 301 includes a CD-ROM drive 3012, an MPU 3013, an MPU 3013, a bus 3014, a ROM 3015, a RAM 3016, and a hard disk 3017. The ROM 3015 stores programs such as a bootup program. The RAM 3016 is connected to the MPU 3013 and temporarily stores application program instructions and provides a temporary storage space. The hard disk 3017 normally stores application programs, system programs, and data. Although not shown here, the computer 301 may further include a network card that provides connection to a LAN.

  A program that causes the computer system 300 to execute the functions of the sales prediction apparatus and the like of the above-described embodiment may be stored in the CD-ROM 3101, inserted into the CD-ROM drive 3012, and further transferred to the hard disk 3017. Further, the program may be transmitted to the computer 301 via a network (not shown) and stored in the hard disk 3017. The program is loaded into the RAM 3016 at the time of execution. The program may be loaded directly from the CD-ROM 3101 or the network.

  The program does not necessarily include an operating system (OS), a third-party program, or the like that causes the computer 301 to execute functions such as the sales prediction apparatus according to the above-described embodiment. The program need only include the part of the instruction that calls the appropriate module in a controlled manner and achieves the desired result. How the computer system 300 operates is well known and will not be described in detail.

  Further, the computer that executes the program may be singular or plural. That is, centralized processing may be performed, or distributed processing may be performed.

  Further, in each of the above embodiments, it goes without saying that two or more communication units existing in one apparatus may be physically realized by one medium.

  In each of the above embodiments, each process (each function) may be realized by centralized processing by a single device (system), or by distributed processing by a plurality of devices. May be.

  The present invention is not limited to the above-described embodiments, and various modifications are possible, and it goes without saying that these are also included in the scope of the present invention.

  As described above, the sales prediction device according to the present invention has an effect that the sales prediction of the product can be performed using an image such as a photograph of the product, and is useful as a sales prediction device.

A Sales prediction device B Learning device 1, 5 Storage unit 2 Reception unit 3, 6 Processing unit 4 Output unit 11 Learning information storage unit 12 Corresponding information storage unit 21 Image reception unit 31 Sales prediction unit 51 Image storage unit 52 Sales performance information storage Unit 61 Learning unit 62 Storage unit 311 First unit 312 Second unit

Claims (8)

A learning information storage unit that stores one or more pieces of learning information including learning image information that is information related to a learning image that is an image of a product and sales performance information that relates to the sales performance of the product;
An image accepting unit that accepts one or more images of a product for sale prediction;
Analyzing one or more images received by the image receiving unit, applying the analysis result to the one or more learning information, and sales prediction information relating to sales prediction of a product related to the image received by the image receiving unit, A sales prediction unit that is acquired using one or more sales performance information included in one or more learning information;
A sales prediction apparatus comprising: an output unit that outputs the sales prediction information. A correspondence information storage unit that stores correspondence information for associating the first sales prediction information with the second sales prediction information;
The sales forecasting unit
The image received by the image receiving unit is analyzed, the analysis result is applied to the one or more learning information, and the first sales prediction is the same kind of information as the one or more sales performance information included in the one or more learning information. A first means of obtaining information;
A second means for obtaining second sales forecast information using the correspondence information from the first sales forecast information,
The output unit is
The sales prediction apparatus according to claim 1, wherein the second sales prediction information is output. The learning information is
The sales prediction device according to claim 1 or 2, which is information learned by using deep learning from learning source information having a learning image and sales performance information. The sales prediction device according to any one of claims 1 to 3, wherein the learning image and the image received by the image receiving unit are photographs of one or more products. The image received by the image receiving unit includes images of two or more products,
The sales forecasting unit
Two or more partial images that are images of each of the two or more products are acquired from the image received by the image receiving unit, the two or more partial images are analyzed, and the analysis result is obtained for each partial image. Applying to the above learning information, obtaining the sales forecast information of the product relating to the image received by the image receiving unit using one or more sales performance information included in the one or more learning information,
The output unit is
The sales prediction device according to any one of claims 1 to 4, wherein the sales prediction information of each of the two or more products is output. The learning information storage unit
One or more pieces of learning information having learning image information and sales performance information that are two or more attribute values of the learning image acquired from the learning image are stored,
The sales forecasting unit
Analyzing the image received by the image receiving unit, obtaining an analysis result that is two or more attribute values, applying the analysis result to the one or more learning information, and a product related to the image received by the image receiving unit The sales prediction device according to any one of claims 1 to 5, wherein the sales prediction information related to the sales prediction is acquired using one or more sales performance information included in the one or more learning information. The recording medium is
A learning information storage unit that stores one or more pieces of learning information including learning image information that is information related to a learning image that is an image of a product and sales result information that relates to the sales result of the product;
A sales prediction method realized by an image reception unit, a sales prediction unit, and an output unit,
An image receiving step in which the image receiving unit receives an image of a product to be sold;
The sales prediction unit analyzes the image received in the image reception step, applies the analysis result to the one or more learning information, and the sales prediction related to the sales prediction of the product related to the image received by the image reception unit. A sales prediction step of acquiring information using one or more sales performance information included in the one or more learning information;
A sales prediction method, wherein the output unit includes an output step of outputting the sales prediction information. Computer-accessible recording media
A learning information storage unit that stores one or more pieces of learning information including learning image information that is information related to a learning image that is an image of a product and sales result information that relates to the sales result of the product;
Computer
An image reception unit for receiving images of products for sale;
The image received by the image receiving unit is analyzed, the analysis result is applied to the one or more learning information, and the sales prediction information related to the sales prediction of the product related to the image received by the image receiving unit is A sales prediction unit that is acquired using one or more sales performance information included in the learning information;
A program for functioning as an output unit for outputting the sales forecast information.