A method for supporting one product with multiple codes

Abstract

The present invention discloses a one-product-multiple-code scenario support method, which relates to the field of information technology, including collecting different codes of the same product and their multi-dimensional attribute data and user behavior data through an interface, calculating the similarity between users based on the user behavior data, standardizing all collected data to obtain a fused data set, using a graph analysis method to select scoring core dimensions from the fused data set and assign initial weights, building a product scoring model to obtain a comprehensive score for each product code, sorting the product codes at the same time, optimizing the comprehensive scores of the product codes, combining the similarity between users to obtain the weights of the optimized core dimensions, and updating the sorting of the product codes; the present invention not only improves the accuracy and stability of the product scoring model through the application of a machine learning algorithm, but also can respond to market changes in real time and dynamically adjust product priorities.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for supporting a one-product-multiple-code scenario. Background Art

[0002] With the rapid development of e-commerce and retail industries, commodity coding management has become increasingly complex. Traditional commodity management usually uses a single coding method to identify commodities. However, in actual applications, the same commodity may have multiple different codes (such as UPC, EAN, internal code, etc.). This phenomenon is called "one product, multiple codes". Especially in the context of global supply chains and cross-platform sales, the "one product, multiple codes" problem has become more and more prominent, bringing many challenges to the company's inventory management, distribution management and data analysis.

[0003] Existing scoring models often rely on fixed weight distribution and cannot flexibly adapt to the needs of different business scenarios. In addition, these scoring models lack a dynamic adjustment mechanism and are difficult to respond to market changes or changes in inventory status in real time, resulting in the possibility that the recommended products may not be the best choice. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for supporting a one-product-multiple-code scenario to solve the problem that the scoring model relies on a fixed weight distribution and cannot flexibly adapt to the needs of different business scenarios.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In a first aspect, the present invention provides a method for supporting a one-product-multiple-code scenario, which comprises:

[0008] Through the interface, different codes of the same product and its multi-dimensional attribute data and user behavior data are collected, and the similarity between users is calculated based on the user behavior data;

[0009] All collected data are standardized to obtain a fused data set;

[0010] Use graph analysis to select the core scoring dimensions from the fused data set and assign initial weights, build a product scoring model to obtain the comprehensive score of each product code, and sort the product codes at the same time;

[0011] Optimize the comprehensive score of the product code, combine it with the similarity between users, obtain the weight of the optimized core dimension, and update the ranking of the products;

[0012] According to the updated sorting of commodity codes, the mapping relationship of commodity codes is adjusted in real time;

[0013] Formulate multi-scenario priority rules and apply them to the functional modules based on the priority score of the product code. As a preferred solution of the one-product-multiple-code scenario support method of the present invention, the different codes of the same product include UPC, EAN, and internal code, the multi-dimensional attribute data includes inventory status, sales status, saleable status, shelf status, geographic location, historical sales and product classification, and the user behavior data includes user ID, purchase history, browsing history, click behavior, dwell time and behavior occurrence time;

[0014] Use user behavior data to build a user-item matrix, and use collaborative filtering algorithm based on the user-item matrix to obtain the similarity between users.

[0015] As a preferred solution of the method for supporting one product with multiple codes in the present invention, wherein: all collected data are standardized to obtain a fused data set, specifically including the following steps:

[0016] The standardization process refers to unifying the different codes of commodities and the formats, units and fields of the corresponding multi-dimensional attribute data, and performing data cleaning on user behavior data;

[0017] The standardized data are merged into the central database to form a fused data set.

[0018] As a preferred solution of the method for supporting the scenario of one product with multiple codes according to the present invention, the graph analysis method is used to select the core dimensions of the score from the fusion data set and assign the initial weights, which specifically includes the following steps:

[0019] Set each data in the fused dataset as a node and use the Pearson correlation coefficient to construct the edges between the nodes;

[0020] Assign an initial importance score to each node, use the PageRank algorithm to iteratively update the importance score of each node, and use the Louvain algorithm to perform community detection, and select the node with the highest PageRank score from each community as the core dimension;

[0021] Assign appropriate weights to each core dimension.

[0022] As a preferred solution of the method for supporting the scenario of one product with multiple codes according to the present invention, a product scoring model is constructed to obtain a comprehensive score for each product code, and the product codes are sorted at the same time, which specifically includes the following steps:

[0023] Calculate the score of each core dimension based on the obtained core dimensions;

[0024] Based on the obtained core dimension scores and corresponding weights, a linear combination method is used to obtain a comprehensive score for the product code.

[0025] According to the calculated comprehensive score of each product code, the products are sorted from high to low, and the products corresponding to the product codes with the highest comprehensive scores are selected for priority display.

[0026] As a preferred solution of the method for supporting the scenario of one product with multiple codes according to the present invention, the comprehensive score of the product code is optimized to obtain the weight of the optimized core dimension and update the ranking of the product code, which specifically includes the following steps:

[0027] Identify sales as a business metric and collect sales of products for each product code;

[0028] The comprehensive score based on the product code uses a linear regression algorithm to predict the business indicators of each product code.

[0029] And calculate the error between the actual business indicator and the predicted business indicator, and apply the activation function to make the error non-negative;

[0030] Based on the similarity between users, predict the user's interest score for products they have never seen;

[0031] According to the error between the actual business indicators and the predicted business indicators and the user's interest score, the learning rate is introduced to control the adjustment range and adjust the weights of the core dimensions;

[0032] Based on the core dimensions with adjusted weights, the comprehensive score of each product code is recalculated, and the sorting is updated according to the comprehensive score to obtain a personalized product score display list.

[0033] As a preferred solution of the method for supporting the one-product-multiple-code scenario of the present invention, wherein: according to the updated sorting of the commodity codes, the normalized commodity mapping relationship is adjusted in real time, specifically comprising the following steps:

[0034] Display the products corresponding to the updated product codes in order, and monitor the changes in the core dimensions of the product codes;

[0035] When a change in the core dimension is detected, the core dimension of the product code is updated immediately, and the comprehensive score of the product code is recalculated using the latest core dimension, and the product code is reordered;

[0036] The products corresponding to the reordered product codes are updated in real time to the displayed product list.

[0037] As a preferred solution of the one-product-multiple-code scenario support method of the present invention, wherein: formulating multi-scenario priority rules and applying them to the functional modules based on the priority scores of the commodity codes, specifically includes the following steps:

[0038] Formulate multi-scenario priority rules based on business needs, build product rating models applicable to multiple scenarios, and apply them to functional modules;

[0039] According to the function module selected by the user, the corresponding product rating model is called to obtain the product display result pointed to by the corresponding product code;

[0040] The product rating model applicable to multiple scenarios has separate weight settings.

[0041] In a second aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the method for supporting a one-product-multiple-code scenario as described in the first aspect of the present invention is implemented.

[0042] In a third aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the method for supporting a one-product-multiple-code scenario as described in the first aspect of the present invention is implemented.

[0043] The beneficial effects of the present invention are as follows: the weights in the product rating model are optimized by using a machine learning algorithm, and the weights of the core dimensions are adjusted based on the error between the actual business indicator sales and the predicted value, thereby achieving self-optimization and continuous improvement of the product rating model. The application of the machine learning algorithm not only improves the accuracy and stability of the product rating model, but also responds to market changes in real time and dynamically adjusts product priorities. This adaptive mechanism ensures that the recommended products are always the best choice, maximizes user needs, and improves the competitiveness and operational efficiency of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0045] Figure 1 This is a flowchart of the method for supporting multiple codes for one product scenario in Example 1.

[0046] Figure 2 This is a schematic diagram of the product rating model in Example 1. DETAILED DESCRIPTION

[0047] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0048] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0049] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0050] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides a method for supporting a one-product-multiple-code scenario, comprising the following steps:

[0051] S1. Collect different codes of the same product and its multi-dimensional attribute data and user behavior data through the interface, and calculate the similarity between users based on the user behavior data.

[0052] The specific steps include:

[0053] Use the API interface to obtain different codes (UPC, EAN, internal code) of the same product in third-party ERP and warehouse management modules and their related multi-dimensional attribute data (inventory status, sales status, saleable status, shelf status, geographic location and historical sales); collect user behavior data (user ID, purchase history, browsing history, click behavior, dwell time and behavior occurrence time) from the user interaction platform.

[0054] Further explanation: Importing data in batches through the API interface enables centralized management of product information with different codes, solving the problem of easy errors in traditional manual entry. This method ensures the consistency and accuracy of all product codes and their related attribute data, avoiding operational errors caused by inconsistent data. The comprehensive collection of multi-dimensional attribute data provides a rich information foundation, which helps to evaluate product performance more scientifically. For example, inventory status and sales status can help companies better plan replenishment strategies. In addition, collecting user behavior data can enable more accurate recommendations, which not only improves the user's shopping experience, but also increases the user's chances of discovering their favorite products, thereby improving user satisfaction and loyalty.

[0055] S1.1. Use user behavior data to build a user-product matrix. This matrix can accurately record each user's specific interaction with different products (such as number of views, purchase frequency, etc.). The user-product matrix is ​​a two-dimensional matrix, in which rows represent users and columns represent products. Each element in the matrix represents the user's interaction score with the product. Traverse all user behavior data and fill in the user-product matrix according to user ID and product code. If the user With goods If there is interaction, ; represents the interaction score, is a specific interaction score (such as number of views, purchase rating, etc.). If there is no interaction, then ;

[0056] For each user, the average interaction score of all interactive items is calculated, and then the collaborative filtering algorithm is used to obtain the similarity between users, which is expressed as:

[0057] ;

[0058] in, Indicates user and users The similarity between Indicates user The interaction score for the product. Indicates user The average engagement score for the items that have been interacted with, Indicates user The interaction score for the product. Indicates user The average engagement score for items that have been interacted with.

[0059] To further illustrate, by calculating the similarity between users, we can find other users whose interests match the target user, and then recommend products that these similar users like. This recommendation method based on user behavior is more accurate than traditional rule engines or simple classification recommendations.

[0060] S2. Standardize all collected data to obtain a fused data set.

[0061] The specific steps include:

[0062] S2.1. Unify the different codes of commodities and the formats, units and fields of the corresponding multi-dimensional attribute data.

[0063] Use ETL tools to define the standard format of each coding type (such as UPC, EAN, internal code), for example, UPC should be 12 digits, EAN should be 13 digits; for multi-dimensional attribute data involving numerical values ​​(such as inventory status, historical sales), clarify its standard unit, for example, inventory status is in "pieces" and historical sales are in "pieces / day"; determine the standard field name and data type of multi-dimensional attribute data. For example, the "inventory status" field is named inventory_status and the data type is integer. ETL tools can handle issues such as data format unification, unit conversion, and field mapping.

[0064] Remove outliers in user behavior data (such as abnormally high dwell time or unreasonable purchase frequency), and then handle missing values ​​in user behavior data by filling the mean.

[0065] All standardized data are stored in a central database in a unified format, unit, and field to form a fused data set.

[0066] It is further explained that the standardized fusion data set provides high-quality basic data for the subsequent product rating model construction. Accurate data is the key to the success of the model, and standardization can significantly improve the prediction accuracy and reliability of the model.

[0067] S3. Use graph analysis to select core scoring dimensions from the fused data set and assign initial weights, build a product scoring model to obtain a comprehensive score for each product code, and sort the product codes at the same time.

[0068] The specific steps include:

[0069] S3.1. Set each multi-dimensional attribute data in the fused data set as a node. For example, inventory status, sales status, and saleable status are all different nodes. Each node can be represented as a vector containing the specific value of the attribute. For example, inventory status can be an integer, sales status can be a Boolean value, etc.

[0070] The Pearson correlation coefficient in the collaborative filtering algorithm is used to calculate the correlation between two attributes, and a correlation threshold (for example, 0.5) is set. If the correlation between the two attributes is greater than the correlation threshold, an edge is established between the two attributes.

[0071] Each node is then assigned an initial importance score, usually set to 1, to ensure fairness in subsequent calculations. The importance score of each node is iteratively updated using the PageRank algorithm

[0072] ;

[0073] in, Representation Node The PageRank score, or importance score, represents the damping factor (usually set to 0.85), Indicates that the node The set of all nodes of Representation Node The PageRank score, Represents a slave node The number of outgoing edges;

[0074] The PageRank algorithm can evaluate the importance of each node in the entire graph, thereby determining which attributes are more influential in the whole. Then the Louvain algorithm is used to maximize the modularity to divide the community. Through community detection, natural groupings between attributes can be found, which helps to understand the inherent structure of the data. The node with the highest PageRank score is selected from each community as the core dimension to ensure that important attributes of different communities are included. This ensures the diversity and representativeness of the core dimensions.

[0075] S3.2. Calculate the score of the core dimension. For example, if the dynamic sales status and historical sales weight are the core dimensions, the score is calculated using a statistical method. The time window is set to the average sales volume per day within 7 days of the week, and the historical sales weight The time window is set to the average sales volume of the past 30 days, and its expression is:

[0076] ;

[0077] ;

[0078] in, The rating value indicating the sales status. represents the sales volume for each day of the week. The score value representing the weight of historical sales volume. represents the sales volume per day in a month. An index representing the day.

[0079] If the saleable status and shelf status of the product code are the core dimensions, the Boolean judgment method is used to obtain the score value. Taking the saleable status as an example, the expression is:

[0080] ;

[0081] in, The rating value indicating the sale status is a Boolean value, which is 1 if the product is available for sale, otherwise 0.

[0082] Set initial weights for the selected core nodes. For example, if inventory status and sales status are selected as core dimensions and inventory is prioritized, the weights of the core dimensions can be set as follows: inventory status weight is 0.7, and sales status weight is 0.3.

[0083] Define and use The calculated core dimension scores are summarized and used to calculate the comprehensive score of the product code. Indicates the index of the core dimension, that is, the number of core dimensions.

[0084] S3.3. Based on the obtained core dimension scores and corresponding weights, a linear combination method is used to obtain the comprehensive score of the product code, which is expressed as follows:

[0085] ;

[0086] in, Indicates the comprehensive score of the product code. Indicates the score value of the core dimension, represents the weight of the core dimension, represents the index of the core dimension, Indicates The rating values ​​of the core dimensions are Indicates The weights of the core dimensions;

[0087] According to the calculated comprehensive score of each product code, the codes are sorted from high to low, and the products corresponding to the product codes with the highest comprehensive scores are selected and displayed to users first. It is directly used as the basis for deciding which products should be displayed first.

[0088] Further explanation: By building a product rating model, a quantitative evaluation of the multi-dimensional performance of products is achieved, providing a scientific basis for decision-making for enterprises. For example, based on the comprehensive rating, enterprises can arrange promotional activities more reasonably and optimize inventory management strategies.

[0089] S4. Optimize the comprehensive score of the product code, and combine it with the similarity between users to obtain the weight of the optimized core dimension, and update the ranking of the product code.

[0090] The specific steps include:

[0091] Select Sales As a key business indicator, since it is a key indicator of product performance. Extract historical sales data for each product code from your ERP. Make sure to cover a long enough period (such as the past few months) to capture seasonality and trend changes.

[0092] S4.1. Taking the comprehensive score of the product code as the independent variable and sales as the dependent variable, a linear regression model is established to obtain the predicted business indicators for each product code. , whose expression is:

[0093] ;

[0094] in, Represents the forecast business indicator of the product code, that is, the forecast sales volume, represents the intercept term, represents the regression coefficient.

[0095] It is further explained that by building a linear regression model, the sales of each product code can be effectively predicted, thus better understanding the performance of the products.

[0096] Forecasted business indicators by product code , calculate actual business indicators and predictive business indicators The error between them is calculated by applying the activation function to make the error non-negative, so as to avoid the negative value affecting the weight adjustment process. The expression is:

[0097] ;

[0098] ;

[0099] in, Represents the error between the actual business indicator and the predicted business indicator, represents the total number of samples, represents the sample index, Indicates The actual business indicators of samples, Indicates The predicted business indicators of samples, Indicates the result of applying the activation function. It indicates the larger value of 0 and the error between the actual business indicator and the predicted business indicator;

[0100] S4.2. Based on the similarity between users , predict the user's interest score for products that they have not touched, and its expression is:

[0101] ;

[0102] in, Indicates the product Interest score, represents the interaction score;

[0103] According to the user's interest score and the error between the actual business indicators and the predicted business indicators, the learning rate is introduced to control the adjustment range to complete the adjustment of the core dimension weights. The expression is:

[0104] ;

[0105] in, Represents the optimized The weights of the core dimensions, represents the learning rate, represents the total number of core dimensions, Represents each weight in all core dimensions, where Represents an index variable, the value range is from 1 to , Indicates user About Products Interest score, Indicates user For The interest scores of products associated with the core dimensions.

[0106] Further explanation: According to the error between the actual business indicators and the predicted business indicators, the weights of each core dimension are dynamically adjusted, so that the product rating model can continuously optimize itself. This method can adapt to changes in the market and business needs, and maintain the timeliness and accuracy of the product rating model.

[0107] Based on the core dimensions with adjusted weights, the comprehensive score of each product code is recalculated, the ranking of the product codes is updated according to the new comprehensive score, and the ranking of the products is adjusted according to the updated product codes to obtain a personalized product rating display list that suits the user's interests.

[0108] Further explanation: According to the error between the actual business indicators and the predicted business indicators and the user's interest rating, the weight of each core dimension is dynamically adjusted, so that the product rating model can continuously optimize itself. This method can adapt to market changes and changes in business needs, meet user interest preferences, and maintain the timeliness and accuracy of the product rating model.

[0109] S5. According to the updated order of the commodity codes, the mapping relationship of the commodities is adjusted in real time.

[0110] The specific steps include:

[0111] After recalculating the comprehensive score of each product code according to the optimized weight, sort the product codes from high to low according to their comprehensive scores. On the front-end display page, display the products corresponding to these product codes to users in the new sorting order. For example, on an e-commerce platform, display the products with the highest scores first.

[0112] Use database triggers to check whether the core dimensions of product codes (such as inventory status, sales status, saleable status, etc.) have changed.

[0113] When a change is detected in the core dimension of a product code, the core dimension data of the product code is updated immediately, and the comprehensive score of the product code is recalculated using the latest core dimension data. For example, if the inventory status of a product changes from in stock to out of stock, its comprehensive score of the product code needs to be re-evaluated. All products are re-sorted based on the new comprehensive score.

[0114] Use WebSocket to update the product list corresponding to the re-ordered product code to the front-end display page in real time, ensuring that the front-end page can respond to back-end data changes immediately, and users can see the latest product recommendations without having to manually refresh the page.

[0115] It is further explained that by recalculating the comprehensive score based on the latest core dimension data, enterprises can obtain a more accurate evaluation of product performance. This provides enterprises with a scientific basis for decision-making, helping them make smarter choices in resource allocation, promotional activities, etc. In addition, the automated monitoring and update mechanism reduces the need for manual intervention and improves operational efficiency. For example, when the inventory status of a certain product changes, its comprehensive score and ranking can be automatically adjusted, avoiding information lag caused by human negligence. Real-time adjustment of the mapping relationship of products can quickly respond to market changes and enable users to see the latest and most popular product recommendations, improving the shopping experience. Especially during promotional activities or when inventory is tight, users can obtain the required product information more quickly, increasing the possibility of purchase.

[0116] S6. Develop multi-scenario priority rules and apply them to functional modules based on the priority scores of product codes.

[0117] The specific steps include:

[0118] S6.1. Develop corresponding priority rules based on the needs of different business scenarios to ensure that products can be effectively evaluated and displayed in each scenario.

[0119] First identify business scenarios, such as distribution management, inventory management, and mini-program tasks. Each scenario has its own specific goals and concerns.

[0120] The focus of distribution management is to optimize supply chain efficiency and improve channel partners' cooperation satisfaction.

[0121] Inventory management focuses on inventory turnover and avoiding overstocking to ensure reasonable inventory levels.

[0122] Mini program tasks focus on user experience and improve user stickiness and conversion rates through personalized recommendations.

[0123] Set specific priority rules for each scenario. For example, in distribution management, you may pay more attention to sales status and historical sales; while in inventory management, you may pay more attention to real-time inventory status.

[0124] S6.2. Build a product rating model suitable for multiple scenarios and apply it to functional modules.

[0125] For example, the inventory management model focuses on the inventory status. The weight of inventory status is set to 0.5, the weight of active sales status is set to 0.15, the weight of available status is set to 0.1, the weight of shelf status is set to 0.1, the weight of geographical location is set to 0.05, and the weight of historical sales is set to 0.1.

[0126] Integrate product rating models in different scenarios into corresponding functional modules, such as inventory management.

[0127] When the user selects a certain function module, the corresponding product rating model is automatically called for calculation and generates the display results of the products corresponding to the corresponding product code.

[0128] It is further explained that by building a dedicated product rating model for different business scenarios, a more accurate product performance evaluation can be obtained. For example, in distribution management, the supply chain strategy can be optimized based on the sales status and historical sales; in inventory management, the inventory layout can be reasonably arranged according to the real-time inventory status. And based on the functional modules selected by the user, the most relevant product recommendations are displayed, which improves the user's shopping experience. For example, in the mini program tasks, products with good sales status and in line with their preferences are recommended to users, which improves user satisfaction and conversion rate.

[0129] This embodiment also provides a computer device, which is applicable to the case of a method for supporting multiple codes for a single product scenario, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the method for supporting multiple codes for a single product scenario proposed in the above embodiment.

[0130] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0131] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for supporting a one-product-multiple-code scenario proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (Static Random Access Memory, SRAM for short), an electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, EEPROM for short), an erasable programmable read-only memory (Erasable Programmable Read Only Memory, EPROM for short), a programmable read-only memory (Programmable Red-Only Memory, PROM for short), a read-only memory (Read-Only Memory, ROM for short), a magnetic storage device, a flash memory, a disk or an optical disk.

[0132] In summary, the present invention achieves self-optimization and continuous improvement of the product rating model by: optimizing the weights in the product rating model using a machine learning algorithm, adjusting the weights of the core dimensions based on the error between the actual business indicator sales and the predicted value. The application of the machine learning algorithm not only improves the accuracy and stability of the product rating model, but also responds to market changes in real time and dynamically adjusts product priorities. This adaptive mechanism ensures that the recommended products are always the best choice, maximizes user needs, and improves the competitiveness and operational efficiency of the enterprise.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for supporting a one-product-multiple-code scenario, characterized in that: include, Through the interface, different codes of the same product and its multi-dimensional attribute data and user behavior data are collected, and the similarity between users is calculated based on the user behavior data; All collected data are standardized to obtain a fused data set; Use graph analysis to select the core scoring dimensions from the fused data set and assign initial weights, build a product scoring model to obtain the comprehensive score of each product code, and sort the product codes at the same time; Optimize the comprehensive score of the product code, combine it with the similarity between users, obtain the weight of the optimized core dimension, and update the ranking of the product code; According to the updated sorting of commodity codes, the mapping relationship of commodities is adjusted in real time; Develop multi-scenario priority rules and apply them to functional modules based on the priority scores of product codes; Optimize the comprehensive score of the product code, obtain the optimized weight of the core dimension, and update the ranking of the product code. The specific steps include: Identify sales as a business metric and collect sales of products for each product code; The comprehensive score based on the product code uses a linear regression algorithm to predict the business indicators of each product code. And calculate the error between the actual business indicator and the predicted business indicator, and apply the activation function to make the error non-negative; Based on the similarity between users, predict the user's interest score for products they have never seen; According to the error between the actual business indicators and the predicted business indicators and the user's interest score, the learning rate is introduced to control the adjustment range and adjust the weights of the core dimensions; Based on the core dimensions with adjusted weights, the comprehensive score of each product code is recalculated, and the sorting is updated according to the comprehensive score to obtain a personalized product score display list.

2. The method for supporting multiple codes for one product scenario according to claim 1, characterized in that: The different codes of the same product include UPC, EAN, and internal code; the multi-dimensional attribute data include inventory status, sales status, saleable status, shelf status, geographic location, historical sales volume, and product classification; the user behavior data includes user ID, purchase history, browsing history, click behavior, dwell time, and behavior occurrence time; Use user behavior data to build a user-item matrix, and use collaborative filtering algorithm based on the user-item matrix to obtain the similarity between users.

3. The method for supporting multiple codes for one product scenario as claimed in claim 2, characterized in that: All collected data are standardized to obtain a fused data set, which includes the following steps: The standardization process refers to unifying the different codes of commodities and the formats, units and fields of the corresponding multi-dimensional attribute data, and performing data cleaning on user behavior data; The standardized data are merged into the central database to form a fused data set.

4. The method for supporting a one-product-multiple-code scenario as claimed in claim 3, characterized in that: The graph analysis method is used to select the core scoring dimensions from the fusion data set and assign initial weights, which includes the following steps: Each multi-dimensional attribute data in the fused data set is set as a node, and the edges between nodes are constructed using the Pearson correlation coefficient; Assign an initial importance score to each node, use the PageRank algorithm to iteratively update the importance score of each node, and use the Louvain algorithm to perform community detection, and select the node with the highest PageRank score from each community as the core dimension; Assign appropriate weights to each core dimension.

5. The method for supporting a one-product-multiple-code scenario as claimed in claim 4, characterized in that: Construct a product rating model to obtain a comprehensive score for each product code and sort the product codes. The specific steps include: Calculate the score of each core dimension based on the obtained core dimensions; Based on the obtained core dimension scores and corresponding weights, a linear combination method is used to obtain a comprehensive score for the product code. According to the calculated comprehensive score of each product code, the products are sorted from high to low, and the products corresponding to the product codes with the highest comprehensive scores are selected for priority display.

6. The method for supporting a one-product-multiple-code scenario as claimed in claim 5, characterized in that: According to the updated sorting of commodity codes, the normalized commodity mapping relationship is adjusted in real time, which specifically includes the following steps: Display the products corresponding to the updated product codes in order, and monitor the changes in the core dimensions of the products; When a change in the core dimension is detected, the core dimension of the product code is updated immediately, and the comprehensive score of the product code is recalculated using the latest core dimension, and the product code is reordered; The products corresponding to the reordered product codes are updated in real time to the displayed product list.

7. The method for supporting one product with multiple codes as claimed in claim 6, characterized in that: Formulate multi-scenario priority rules and apply them to functional modules based on the priority scores of product codes. The specific steps include the following: Formulate multi-scenario priority rules based on business needs, build product rating models applicable to multiple scenarios, and apply them to functional modules; According to the function module selected by the user, the corresponding product rating model is called to obtain the product display result pointed to by the corresponding product code; The product rating model applicable to multiple scenarios has separate weight settings.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for supporting a one-product-multiple-code scenario described in any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for supporting a one-product-multiple-code scenario described in any one of claims 1 to 7 are implemented.

Images