Information processing device, information processing method, and information processing program

Abstract

To provide support in recommending products to customers. [Solution] The information processing device 100 includes an acquisition means for acquiring data on customer purchase history and data on store management as a training dataset, and a learning means for training a machine learning model to output data indicating a recommendation engine to be used to recommend products to customers when policy data indicating the content of ongoing sales promotion measures is input using the training dataset acquired by the acquisition means, and an input means for inputting policy data indicating the content of ongoing sales promotion measures into the trained machine learning model, and a recommendation means for executing a process to recommend products to customers using the recommendation engine when data indicating a recommendation engine to be used to recommend products to customers is output from the trained machine learning model.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and an information processing program.

Background Art

[0002] A recommendation engine that provides information for arousing the purchasing desire of products is known (for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In general, the logic of a recommendation engine used to recommend products to customers varies depending on what is emphasized when selecting products. Therefore, if recommendation engines with different logics are used, the products recommended to customers may change. For this reason, when there are multiple recommendation engines with different logics, a mechanism for supporting the selection of the recommendation engine to be used is required, but conventionally, no technology for this has been studied.

Means for Solving the Problems

[0005] The information processing apparatus according to the present invention This includes information on products the customer has purchased in the past and the recommendation engine used to suggest those products to the customer. acquisition means for acquiring, as a learning data set, data of purchase history information regarding a customer's purchase history and This includes policy information that shows the details of sales promotion measures implemented in the past. data of management information regarding the management of a store, and when inputting measure data indicating the content of measures for promoting sales in progress using the learning data set acquired by the acquisition means, Data showing the recommendation engines used when the same measures were implemented in the past, resulting in customers purchasing products. data indicating a recommendation engine to be used for recommending products to customers asA learning method for training a machine learning model to produce an output. The system includes an input means for inputting policy data indicating the content of ongoing sales promotion measures into a trained machine learning model trained by a learning means, and a recommendation means for executing a process to recommend products to customers using a recommendation engine when data indicating a recommendation engine to be used to recommend products to customers is output from the trained machine learning model. Features that 。 Book The information processing method according to the invention is characterized by an acquisition means, This includes information on products the customer has purchased in the past and the recommendation engine used to suggest those products to the customer. Customer purchase history data and This includes policy information that shows the details of sales promotion measures implemented in the past. The process involves acquiring management information data related to store operations as a training dataset, and then the learning method uses the acquired training dataset to input data indicating the content of ongoing sales promotion measures. Data showing the recommendation engines used when the same measures were implemented in the past, resulting in customers purchasing products. Data showing recommendation engines that should be used to recommend products to customers. as Steps to train a machine learning model to produce output and The input means inputs measure data indicating the content of ongoing sales promotion measures into a trained machine learning model trained by the learning means, and the recommendation means, when data indicating a recommendation engine to be used to recommend products to customers is output from the trained machine learning model, executes a process to recommend products to customers using the recommendation engine. A method performed on a computer, which has the following characteristics. 。 Book The information processing program according to the invention is This includes information on products the customer has purchased in the past and the recommendation engine used to suggest those products to the customer. Customer purchase history data and This includes policy information that shows the details of sales promotion measures implemented in the past. The process involves acquiring management information data related to store operations as a training dataset, and then inputting data showing the content of ongoing sales promotion measures using the acquired training dataset. Data showing the recommendation engines used when the same measures were implemented in the past, resulting in customers purchasing products. Data showing recommendation engines that should be used to recommend products to customers. as Training procedure for training a machine learning model to produce output The process involves inputting data indicating the content of ongoing sales promotion measures into a trained machine learning model trained in the learning procedure, and a recommendation procedure that, when the trained machine learning model outputs data indicating a recommendation engine to be used to recommend products to customers, executes a process to recommend products to customers using that recommendation engine. This is a program that causes a computer to execute something. [Effects of the Invention]

[0006] According to the present invention, it is possible to provide a mechanism for supporting the selection of a recommendation engine to be used to recommend products to customers. [Brief explanation of the drawing]

[0007] [Figure 1] This is a block diagram showing the configuration of one embodiment of the information processing device 100. [Figure 2]This is a flowchart illustrating the processing flow during the training phase of a machine learning model. [Figure 3] This is a flowchart illustrating the processing flow using the AI ​​agent in the first embodiment. [Figure 4] This is a flowchart illustrating the processing flow using the AI ​​agent in the second embodiment. [Modes for carrying out the invention]

[0008] —First Embodiment— Figure 1 is a block diagram showing the configuration of one embodiment of the information processing device 100 in this embodiment. The information processing device 100 may be a personal computer or a server, and includes a communication module 101, a control device 102, and a recording device 103.

[0009] The communication module 101 includes a module for connecting the information processing device 100 to a communication line such as a LAN or the Internet, either wirelessly or via a wired connection. In this embodiment, the information processing device 100 can communicate with other devices by connecting to a communication line via the communication module 101.

[0010] The control device 102 consists of a CPU, memory, and other peripheral circuits, and controls the entire information processing device 100. The memory that makes up the control device 102 is a volatile memory such as SDRAM. This memory is used as work memory for the CPU to load programs when executing programs, and as buffer memory for temporarily recording data. For example, data received via the communication module 101 is temporarily recorded in the buffer memory.

[0011] The recording device 103 is a recording device for recording various data stored by the information processing device 100, data of programs to be executed by the control device 102, etc. For example, an HDD (Hard Disk Drive), an SSD (Solid State Drive), etc. are used. Note that the program data recorded in the recording device 103 is provided by being recorded on a recording medium such as a CD-ROM or a DVD-ROM, or is provided via a network, and the control device 102 can execute the program by installing the program data acquired by the user in the recording device 103.

[0012] In the information processing device 100 in the present embodiment, when there are a plurality of recommendation engines with different logics in a situation where a recommendation engine for recommending products sold in a store to customers is used, based on the customer's purchase history and management information regarding the management of the store, an optimal recommendation engine for recommending products to the customer is selected to provide a mechanism for recommending products to the customer. Note that in the present embodiment, the store includes a physical store and a virtual store on the Internet. Also, the products include products sold to customers and service products provided to customers.

[0013] Generally, there are countless recommendation engines with characteristics in their logics as known recommendation engines. For example, as a recommendation engine for recommending cosmetics to customers, there are various recommendation engines with characteristics in their logics, such as a recommendation engine that adopts a logic of proposing recommended products based on the ingredients contained in the products, a recommendation engine that adopts a logic of proposing recommended products based on the prices of the products, a recommendation engine that adopts a logic of proposing bulk purchases of products, or a recommendation engine that adopts a logic of proposing recommended products based on the products the customer has purchased in the past.

[0014] In a situation where there are countless recommendation engines each with unique logic, the products recommended to customers can vary depending on which recommendation engine is adopted. This means that the choice of recommendation engine can influence whether customers actually purchase the recommended products. In view of such a situation, the information processing apparatus 100 in the present embodiment assists in selecting the optimal recommendation engine by determining which recommendation engine should be used to recommend products to customers in a situation where multiple recommendation engines exist. Hereinafter, the processing executed by the information processing apparatus 100 in the present embodiment will be described.

[0015] In the information processing apparatus 100 in the present embodiment, as described above, a mechanism for selecting an optimal recommendation engine for recommending products to customers is provided based on the customer's purchase history and business information related to the store's operation. For this purpose, in the information processing apparatus 100, purchase history information related to the customer's purchase history and business information related to the store's operation are recorded in the recording apparatus 103 in advance. The purchase history information includes, for example, for each customer, information on the purchased product, purchase date, purchase price, and the recommendation engine information used to recommend that product to the customer. The business information includes, for example, measure information indicating the content of measures for past sales promotion.

[0016] Note that in the present embodiment, the measure information indicating the content of measures for sales promotion is information indicating the content of activities for promoting the sale of products, and assumes activities that affect the number of product sales, such as the distribution of leaflets, the distribution of coupons, price discounts, the distribution of pocket tissues, web advertisements, TV commercials, radio commercials, etc. Note that these activities can also be implemented in combination.

[0017] In the information processing device 100, the control device 102 acquires data on purchase history information and management information recorded in the recording device 103 as a training dataset, and inputs the acquired training dataset into a machine learning model. Then, when policy data indicating the content of ongoing sales promotion measures is input, the machine learning model is trained to output data indicating the recommendation engine that should be used to recommend products to customers. In this way, when policy data indicating the content of currently implemented sales promotion measures is input as an explanatory variable, a machine learning model can be created that selects the recommendation engine that should be used to recommend products to customers from among multiple recommendation engines with different logics, and outputs data indicating the selected recommendation engine as the target variable.

[0018] The control device 102 records data for use with the created trained machine learning model. This allows the control device 102 to perform processing using the recorded trained machine learning model. The destination for recording the trained machine learning model is not particularly limited. For example, the control device 102 may record the trained machine learning model in the recording device 103, or it may record it in another device that can communicate via a communication line. In the following description, it will be assumed that the trained machine learning model is recorded in the recording device 103.

[0019] In this embodiment, the information processing device 100 provides an AI agent that operates as a human agent using this machine learning model. In this embodiment, the AI ​​agent uses the trained machine learning model described above to select a recommendation engine to use to recommend products to customers, and then uses the selected recommendation engine to perform the process of recommending products to customers. For this purpose, in this embodiment, a program for realizing the AI ​​agent is recorded in the recording device 103, and the control device 102 executes this program, thereby executing the functions provided by the AI ​​agent.

[0020] Here, an AI agent is a system that acts as a substitute (agent) for a human, possessing the ability to learn and make decisions on its own, and integrating various AI technologies for solving complex problems. For example, when an AI agent is given a question (prompt) by a user, it searches for or generates an answer to the question and outputs the answer to the question. Also, for example, when a task (prompt) is given by a user, it searches for or generates a solution to the task and outputs the solution to the task.

[0021] In this embodiment, an LLM (Large-Scale Language Model) is assumed as an example of an AI agent. However, the AI ​​agent according to the present invention is not limited to an LLM (Large-Scale Language Model).

[0022] In this embodiment, the AI ​​agent inputs policy data, which indicates the content of sales promotion measures currently being implemented, from store staff, into a trained machine learning model. Based on the input policy data, the trained machine learning model outputs data as an inference result indicating the recommendation engine that should be used to recommend products to customers, as described above.

[0023] When the AI ​​agent receives data from a trained machine learning model indicating a recommendation engine to use to recommend products to a customer, it launches the corresponding recommendation engine based on the output data and identifies products to recommend to the customer based on the output from the recommendation engine. The program for running the recommendation engine may be stored in the recording device 103 or in an external device that can be connected via a communication line.

[0024] When the AI ​​agent receives product information from the recommendation engine, it processes the output to recommend the product to the customer. The method of recommending the product to the customer is not particularly limited; for example, the product output from the recommendation engine may be sent as a recommended product to the customer's terminal connected via a communication line. Alternatively, the product output from the recommendation engine may be sent as a recommended product to a store terminal installed in the store or to a store employee's terminal, notifying the employee. In this case, the employee can then verbally recommend the product to the customer.

[0025] Using the information processing device 100 in this embodiment, products can be recommended to customers as follows, for example. For example, suppose that the purchase history information of customer A, learned by a machine learning model, includes a purchase history showing that product B was recommended by a recommendation engine that employs a logic to suggest recommended products based on the price of the products, and as a result, customer A purchased product B. Furthermore, suppose that the management information includes policy information showing that on the day customer A purchased product B, a flyer featuring product B was distributed.

[0026] In this case, when the AI ​​agent inputs policy data indicating that flyers featuring product B are currently being distributed into a trained machine learning model, the trained machine learning model outputs an inference result suggesting the use of a recommendation engine that employs a logic to suggest products based on their price in order to recommend product B to customer A.

[0027] The AI ​​agent activates a recommendation engine that employs a logic to suggest products based on their price, using the output from a trained machine learning model. It then processes the engine to identify products to recommend to the customer and, based on the output from the recommendation engine, processes the engine to recommend the products to the customer.

[0028] This allows us to recommend product B to customer A again, taking into account that customer A previously purchased product B when it was featured in a flyer, using a recommendation engine that uses a price-based recommendation logic. This is based on the information that customer A purchased product B, and considering that a similar promotion is currently underway. As a result, we can increase the likelihood that customer B will purchase the recommended product B.

[0029] Figure 2 is a flowchart showing the processing flow during the training phase of a machine learning model executed by the information processing device 100 in this embodiment. The processing shown in Figure 2 is executed by the control device 102 as a program that is started at the timing when the machine learning model is trained. The timing of training the machine learning model is not particularly limited.

[0030] In step S10, the control device 102 acquires the purchase history data and management information data recorded in the recording device 103 as a training dataset, as described above. Then, the process proceeds to step S20.

[0031] In step S20, the control device 102 inputs the acquired training dataset into the machine learning model to train the machine learning model. After that, the process proceeds to step S30.

[0032] In step S30, the control device 102 determines whether or not the machine learning model has finished training. For example, the control device 102 may determine that the machine learning model has finished training when the operator instructs it to do so. If the determination in step S30 is negative, the process returns to step S10. Conversely, if the determination in step S30 is positive, the process proceeds to step S40.

[0033] In step S40, the control device 102 records the data for using the trained machine learning model to a recording destination, such as the recording device 103, as described above. After that, it terminates the process.

[0034] Figure 3 is a flowchart showing the processing flow using the AI ​​agent executed by the information processing device 100 in the first embodiment. The processing shown in Figure 3 is executed by the control device 102 as a program that is started when the operator instructs the AI ​​agent to be started.

[0035] In step S110, the control device 102 inputs the policy data, which indicates the content of the sales promotion measures currently being implemented and which was entered by the store staff, into the trained machine learning model. Then, the process proceeds to step S120.

[0036] In step S120, the control device 102 determines whether the trained machine learning model has output the inference result data described above, that is, data indicating a recommendation engine that should be used to recommend products to customers. If the determination in step S120 is positive, the process proceeds to step S130.

[0037] In step S130, the control device 102 identifies and activates a recommendation engine to be used to recommend products to customers, based on the inference results output from the trained machine learning model, i.e., data indicating the recommendation engine that should be used to recommend products to customers, as described above. The process then proceeds to step S140.

[0038] In step S140, the control device 102 determines whether or not the recommendation engine has output a recommended product to the customer. If the determination in step S140 is positive, the process proceeds to step S150.

[0039] In step S150, the control device 102 performs the process of recommending products to the customer, as described above. After that, the process ends.

[0040] According to the first embodiment described above, the following effects and advantages can be obtained. (1) The control device 102 acquires data on customer purchase history and data on store management as training datasets. Using the acquired training datasets, it takes data on measures indicating the content of ongoing sales promotion measures as input and trains a machine learning model to output data indicating the recommendation engine that should be used to recommend products to customers. This makes it possible to create a machine learning model that can output the optimal recommendation engine to use to recommend products to customers.

[0041] (2) The machine learning model selects the recommendation engine that should be used to recommend products to customers from among multiple recommendation engines with different logics, and outputs data indicating the selected recommendation engine. This makes it possible to determine and suggest which recommendation engine should be used to recommend products to customers in a situation where multiple recommendation engines with distinctive logics exist.

[0042] (3) Purchase history information includes, for each customer, the purchased items, the purchase date, the purchase price, and information about the recommendation engine used to recommend those items to the customer. This allows machine learning models to learn when, what, and at what price customers purchase items, and which recommendation engine was used to recommend those items.

[0043] (4) The management information now includes policy information that shows the details of sales promotion measures implemented in the past. This allows the machine learning model to learn policy information that shows the details of sales promotion measures implemented in the past.

[0044] (5) The control device 102 takes measure data indicating the content of ongoing sales promotion measures as input and outputs data indicating a recommendation engine that should be used to recommend products to customers as input to a pre-trained machine learning model. When the pre-trained machine learning model outputs data indicating a recommendation engine that should be used to recommend products to customers, the control device 102 takes measure data indicating the content of ongoing sales promotion measures as input and executes processing to recommend products to customers using that recommendation engine. This makes it possible to support the selection of the optimal recommendation engine to use to recommend products to customers.

[0045] (6) When the trained machine learning model outputs data indicating a recommendation engine that should be used to recommend products to the customer, the control device 102 starts the output recommendation engine and executes a process to recommend the products output by the recommendation engine to the customer as recommended products. This makes it possible to execute a process to recommend products to the customer using the optimal recommendation engine.

[0046] —Second Embodiment— In the first embodiment described above, an example was explained in which purchase history data and management information data were input into a machine learning model as training datasets, and when policy data indicating the content of ongoing sales promotion measures was input, the machine learning model was trained to output data indicating a recommendation engine that should be used to recommend products to customers. Furthermore, an example was explained in which, when policy data indicating the content of ongoing sales promotion measures was input to the trained machine learning model, data indicating a recommendation engine that should be used to recommend products to customers was output as an inference result based on the input policy data.

[0047] In contrast, the second embodiment describes an example in which a trained machine learning model outputs data indicating a recommendation engine that should be used to recommend products to a customer, based on learned customer purchase history data, even if there is no policy data indicating the content of ongoing sales promotion measures. In the second embodiment, the contents shown in Figures 1 and 2 are the same as in the first embodiment, so their explanation is omitted.

[0048] In the second embodiment, the trained machine learning model determines the customer's purchasing decision factors based on the learned customer purchase history data, and selects a recommendation engine to use to recommend products to the customer based on the determined purchasing decision factors. Here, the customer's purchasing decision factors are assumed to include price and promotional sensitivity, product attributes and copy preferences, quantity and size preferences, timing and rhythm, store and channel preferences, co-purchase and usage estimation (basket), loyalty and switching, payment, points and coupon behavior, and long-term change and life event indications, as well as supply and shelf influences.

[0049] Price and promotion sensitivity is a purchasing decision factor that represents how customers react to price and promotions. Based on customer purchase history data, a trained machine learning model determines customer price and promotion sensitivity, such as (1) fractional effect and threshold effect, where the purchase rate is low at 200 yen but increases at 198 yen; (2) differences in purchase rates depending on the type of discount, such as actual price discounts, percentage off, point multipliers, and bulk discounts; (3) customer response to discount coupons; (4) the effect of flyers and special sales; (5) multi-buy responses such as "buy 3 for X yen" or "buy 2 half price"; (6) whether regular customers continue or leave due to price revisions; or (6) customer preference for volume discounts. The model then selects a recommendation engine that corresponds to the customer's price and promotion sensitivity.

[0050] Product attributes and copy preferences are purchasing decision factors that indicate what kind of products customers prefer to choose, and what kind of catchphrases they prefer. Based on customer purchase history data, a trained machine learning model determines customer product attributes and copy preferences, such as (1) ingredient and nutrition appeals such as high protein, dietary fiber, caffeine-free, or allergen-free; (2) health claims such as additive-free, organic, low sugar, or reduced salt; (3) flavor and texture words such as rich, full-bodied, refreshing, or moist; (4) brand or sub-line fixedness such as premium, light, or seasonal; (5) raw materials, origin, and manufacturing methods that elicit a response to phrases such as domestic, regional name, or direct-fire roasted; or (6) packaging specifications that elicit a response to convenience such as individual packaging, stand-up pouch, or resealable cap. The model then selects a recommendation engine that matches the product attributes and copy preferences. The trained machine learning model may also be used to detect information registered in the product master by quantifying it in order to select a recommendation engine that matches the product attributes and copy preferences.

[0051] Quantity and size preferences are purchasing decision factors that indicate what quantity or size of product a customer prefers to select. Based on customer purchase history data, a trained machine learning model determines the customer's quantity and size preferences, such as (1) bulk buying habits showing fixed patterns like buying 3 items each time, (2) pantry loading such as buying large quantities during sale weeks and extending the interval until the next purchase, (3) clear preferences for demand units such as single servings or family sizes, (4) whether the customer prioritizes unit price such as price per 100g or price per 100ml, or package price, or (5) a refill / refill orientation such as switching to purchasing only refills after purchasing the main product once, and selects a recommendation engine according to the customer's quantity and size preferences.

[0052] Timing and rhythm are purchasing decision factors that represent the timing or rhythm at which customers purchase products. Based on customer purchase history data, trained machine learning models determine customer timing and rhythm, such as (1) day-of-the-week and time-of-day patterns like weekday evenings or weekend mornings, (2) around payday, or at the beginning or end of the month, (3) seasonality and pre-event patterns like Golden Week, Obon, New Year's, or Halloween, (4) regularity of repeat cycles indicating replenishment cycles like repurchases every n days ± α, or (5) weather-related factors such as improved accuracy due to external weather linkage or improved sales of instant foods on rainy days, and select a recommendation engine that matches the customer's timing and rhythm.

[0053] Store and channel preferences are purchasing decision factors that represent which stores and channels customers prefer to choose. Based on customer purchase history data, trained machine learning models determine customer store and channel preferences such as (1) store loyalty or switching between stores, such as using stores near home or near work; (2) point day orientation, such as visiting stores or increasing purchase rates on days when points are increased; or (3) using online stores and physical stores interchangeably, and select a recommendation engine that matches the store and channel preferences.

[0054] Co-purchase and usage estimation (basket) is a purchasing decision factor that represents what kinds of products customers buy together and the estimated uses of those products. Based on customer purchase history data, a trained machine learning model determines customer co-purchase and usage estimation (basket), such as (1) co-purchase rules like milk and cereal or yogurt and honey, (2) recipe bundles like curry ingredient sets or salad ingredient sets, (3) substitution and complementary relationships like butter or margarine or coffee and milk, (4) life stage indications like diapers and baby wipes or pet food and litter, or (5) health-conscious sets like protein, Greek yogurt and nuts, and selects a recommendation engine that corresponds to the co-purchase and usage estimation (basket).

[0055] Loyalty switching is a purchasing decision factor that represents a customer's loyalty to a product or their willingness to switch products. Based on customer purchase history data, a trained machine learning model determines customer loyalty switching, such as (1) brand / manufacturer loyalty, including a high concentration of market share within a category; (2) switch triggers, including changing brands due to price increases, sales, or the introduction of new products; (3) a willingness to try new products, including a high rate of trial purchases immediately after launch; or (4) signs of churn, such as switching to similar brands due to an extended purchase cycle. The model then selects a recommendation engine that corresponds to the customer's loyalty switching.

[0056] Payment, points, and coupon behavior are purchasing decision factors that represent customer actions related to payment, point usage, or coupon usage. Based on customer purchase history data, a trained machine learning model determines customer payment, points, and coupon behavior, such as (1) whether the lag between coupon acquisition and usage is early use or delayed use, (2) whether the payment method is fixed, such as QR code payment or credit card payment, or (3) whether the customer responds to membership rank or reward rates, such as top members tending to make bulk purchases, and selects a recommendation engine that corresponds to that payment, points, and coupon behavior.

[0057] Long-term change and life event indicators are purchasing decision factors that represent long-term changes and life events in the customer. Based on customer purchase history data, the trained machine learning model determines long-term changes and life event indicators in the customer, such as (1) changes in family structure, including the introduction and graduation of baby products; (2) changes in residence or work location, including geographical shifts in stores used; or (3) changes in health and preferences, such as a gradual shift to low-sugar or low-salt categories, and selects a recommendation engine that corresponds to the long-term change and life event indicators.

[0058] Supply and shelf impact are purchasing decision factors that represent the supply and shelf impact of a product. A trained machine learning model, based on customer purchase history data, determines supply and shelf impacts such as (1) zero sales of the target SKU at that store on that day + substitute purchases by regular customers of nearby SKUs due to stock shortage estimation, or (2) a sharp increase in sales of a specific SKU in the same week at a store = the impact of shelf rearrangement and display enhancements suggesting a policy, and selects a recommendation engine according to the supply and shelf impact. In addition, supply and shelf impacts can be observed vicariously based on ID-POS data.

[0059] The AI ​​agent inputs prompts to instruct the trained machine learning model to select a recommendation engine. This prompts the trained machine learning model to determine the customer's purchasing decision factors, and then, based on those factors, select and output the recommendation engine that should be used to recommend products to the customer.

[0060] The AI ​​agent activates a recommendation engine that employs a logic to suggest products based on the output of a trained machine learning model. It processes the engine to identify products to recommend to the customer and then processes the engine to recommend those products to the customer based on the output from the recommendation engine.

[0061] This allows you to select a recommendation engine that employs a logic to suggest products to customers based on their purchase history, thereby increasing the likelihood that customers will actually purchase the recommended products.

[0062] Figure 4 is a flowchart showing the processing flow using the AI ​​agent executed by the information processing device 100 in the second embodiment. The processing shown in Figure 4 is executed by the control device 102 as a program that is started when the operator instructs the AI ​​agent to be started.

[0063] In step S210, the control device 102 inputs a prompt to the trained machine learning model to instruct it to select a recommendation engine, as described above. The process then proceeds to step S220.

[0064] In step S220, the control device 102 determines whether the trained machine learning model has output the inference result data described above, that is, data indicating a recommendation engine that should be used to recommend products to customers. If the determination in step S220 is positive, the process proceeds to step S230.

[0065] In step S230, the control device 102 identifies and starts a recommendation engine to be used to recommend products to customers, based on the inference results output from the trained machine learning model, i.e., data indicating the recommendation engine that should be used to recommend products to customers, as described above. Then, the process proceeds to step S240.

[0066] In step S240, the control device 102 determines whether or not the recommendation engine has output a recommended product to the customer. If the determination in step S240 is positive, the process proceeds to step S250.

[0067] In step S250, the control device 102 performs the process of recommending products to the customer, as described above. After that, the process ends.

[0068] According to the second embodiment described above, in addition to the effects of the first embodiment described above, the following effects can be obtained. (1) The control device 102 acquires data on customer purchase history and data on store management as training datasets. Using the acquired training datasets, it instructs the control device to select a recommendation engine to use to recommend products to customers, and trains a machine learning model to output data indicating the recommendation engine to use to recommend products to customers. This supports the selection of the optimal recommendation engine to use to recommend products to customers.

[0069] (2) When the control device 102 is instructed to select a recommendation engine to be used to recommend products to a customer, it inputs the instruction to select a recommendation engine to be used to recommend products to a customer to a pre-trained machine learning model that has been trained to output data indicating the recommendation engine to be used to recommend products to a customer. When the pre-trained machine learning model outputs data indicating the recommendation engine to be used to recommend products to a customer, the control device 102 uses that recommendation engine to execute the process of recommending products to the customer. This makes it possible to select a recommendation engine that employs a logic to suggest recommended products to the customer based on the customer's purchase history and recommend products to the customer. As a result, the likelihood of the customer purchasing the recommended product can be increased.

[0070] —Revised Version— Furthermore, the information processing device of the above-described embodiment can also be modified as follows. (1) In the first and second embodiments described above, the purchase history information relating to a customer's purchase history was described in an example that included, for each customer, the purchased product, the purchase date, the purchase price, and information about the recommendation engine used to recommend the product to the customer. However, the purchase history information relating to a customer's purchase history may also include the following information.

[0071] Data obtained from interactions between the LLM (Large-Scale Language Model) and customers may include, for example, information entered by customers for shopping consultations into the LLM embedded in the company's app. Data showing customer behavior on the app may also be included, such as history information on coupon acquisition, campaign participation, and product searches. Furthermore, data on customer shopping routes may be included, such as information identified by tracking customers' movements within the store using cameras, sensors, or estimations based on purchased items. Data on customer shopping behavior may also be included, such as information identified by tracking products customers pick up using cameras or by digitizing their eye movements. Additionally, data on web searches and browsing history may be included, such as information on web product searches and related keyword searches. Finally, unique data such as health data and skin data may be included, for example, in the case of skin data, information showing the correlation between a dry skin diagnosis and the purchase of moisturizing products.

[0072] Furthermore, the present invention is not limited in any way to the configurations in the embodiments described above, as long as the characteristic functions of the present invention are not impaired. [Explanation of Symbols]

[0073] 100 Information Processing Devices 101 Communication Module 102 Control device 103 Recording device

Claims

1. An acquisition means for acquiring, as a training dataset, purchase history data relating to a customer's purchase history, which includes products the customer has purchased in the past and information from a recommendation engine used to recommend those products to the customer, and management information data relating to the operation of a store, which includes information on measures indicating the content of sales promotion measures implemented in the past. A learning means that, using the training dataset acquired by the acquisition means, inputs data indicating the content of ongoing sales promotion measures, and trains a machine learning model to output data indicating the recommendation engine that was used when the same measures were implemented in the past and resulted in customers purchasing products, as data indicating the recommendation engine that should be used to recommend products to customers. An input means for inputting policy data indicating the content of ongoing sales promotion measures into a trained machine learning model trained by the aforementioned learning means, An information processing apparatus comprising: a recommendation means that, when data indicating a recommendation engine to be used to recommend products to a customer is output from the aforementioned trained machine learning model, executes a process to recommend products to a customer using the recommendation engine.

2. In the information processing apparatus according to claim 1, The machine learning model is an information processing device characterized by selecting a recommendation engine from among multiple recommendation engines with different logics that should be used to recommend products to customers, and outputting data indicating the selected recommendation engine.

3. In the information processing apparatus described in Claim 1, The aforementioned recommendation means is an information processing device characterized by activating a recommendation engine when data indicating a recommendation engine to be used to recommend products to a customer is output from the trained machine learning model, and executing a process to recommend the products output from the recommendation engine to the customer as recommended products.

4. In the information processing apparatus according to Claim 1, The information processing device is characterized in that the purchase history information includes at least one of the following: data obtained from a large-scale language model and interaction with the customer, data showing the customer's behavior on the app, data on the customer's shopping flow, data on the customer's shopping behavior, data on web search and browsing history, and customer health data and skin data.

5. The acquisition method involves the steps of acquiring data on customer purchase history, which includes information on products purchased by the customer in the past and the recommendation engine used to recommend those products to the customer, and data on store management, which includes information on measures taken to promote sales in the past, as a training dataset. The learning means, using the training dataset acquired by the acquisition means, takes policy data indicating the content of ongoing sales promotion measures as input, and trains a machine learning model to output data indicating recommendation engines that were used when the same measures were implemented in the past and resulted in customers purchasing products, as data indicating recommendation engines that should be used to recommend products to customers. The input means includes the step of inputting measure data indicating the content of the sales promotion measures being implemented into the trained machine learning model trained by the learning means, A computer-based information processing method comprising the steps of: when the recommendation means outputs data indicating a recommendation engine that should be used to recommend a product to a customer from the trained machine learning model, the method executes a process to recommend a product to a customer using the recommendation engine.

6. In the information processing method described in claim 5, The machine learning model is characterized by selecting a recommendation engine that should be used to recommend products to customers from among multiple recommendation engines with different logics, and outputting data indicating the selected recommendation engine.

7. In the information processing method described in Claim 5, The aforementioned recommendation means is an information processing method characterized by activating a recommendation engine when data indicating a recommendation engine to be used to recommend products to a customer is output from the trained machine learning model, and executing a process to recommend the products output from the recommendation engine to the customer as recommended products.

8. In the information processing method described in Claim 5, The purchase history information is characterized by including at least one of the following: data obtained from a large-scale language model and interaction with the customer, data showing the customer's behavior on the app, data on the customer's shopping flow, data on the customer's shopping behavior, data on web search and browsing history, and customer health data and skin data.

9. A procedure for acquiring a learning dataset comprising: data on customer purchase history, which includes information on products previously purchased by the customer and information on the recommendation engine used to recommend those products to the customer; and data on store management, which includes information on measures indicating the content of sales promotion measures implemented in the past; A learning procedure in which, using the training dataset obtained in the acquisition procedure described above, a machine learning model is trained to output data indicating the recommendation engine that should be used to recommend products to customers, based on data indicating the recommendation engine that was used when the same measure was implemented in the past and resulted in customers purchasing products, when the training data is input, and the training data indicates the recommendation engine that should be used to recommend products to customers. An input procedure is provided to input measure data indicating the content of ongoing sales promotion measures into the trained machine learning model trained using the aforementioned learning procedure, An information processing program that, when data indicating a recommendation engine to be used to recommend products to customers is output from the aforementioned trained machine learning model, causes a computer to execute a recommended procedure for performing the process of recommending products to customers using the recommendation engine.

10. In the information processing program described in claim 9, The aforementioned machine learning model is an information processing program characterized by selecting a recommendation engine that should be used to recommend products to customers from among multiple recommendation engines with different logics, and outputting data indicating the selected recommendation engine.

11. In the information processing program described in Claim 9, The aforementioned recommended procedure is an information processing program characterized by launching a recommendation engine when data indicating a recommendation engine to be used to recommend products to a customer is output from the trained machine learning model, and then executing a process to recommend the products output from the recommendation engine to the customer as recommended products.

12. In the information processing program described in Claim 9, The purchase history information is characterized by including at least one of the following: data obtained from a large-scale language model and interaction with the customer, data showing the customer's behavior on the app, data on the customer's shopping flow, data on the customer's shopping behavior, data on web search and browsing history, and customer health data and skin data.

Images