A product operation management method and system based on knowledge graph

Abstract

The present invention discloses a product operation management method and system based on knowledge graph, which relates to the technical field of product operation management, and specifically includes the following steps: obtaining comprehensive characteristic information of each substitute commodity in real time, and generating function matching coefficient and user acceptance index of each substitute commodity respectively; constructing a substitution evaluation model for the function matching coefficient and user acceptance index of each substitute commodity generated, generating substitution evaluation coefficient of each substitute commodity, evaluating the substitution degree of each substitute commodity for non-purchasable commodities, and dividing each substitute commodity into high-substitutability commodities, potential substitute commodities and low-substitutability commodities according to the evaluation results; and taking different management measures for high-substitutability commodities, potential substitute commodities and low-substitutability commodities according to the division results of each substitute commodity. The present invention solves the problem of inaccurate recommendation of substitute commodities, and realizes accurate recommendation effect based on dynamic optimization of user demand and market characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of product operation management, and in particular to a product operation management method and system based on knowledge graph. Background Art

[0002] Product operation refers to the use of a series of systematic management measures to promote the smooth operation of the entire life cycle of products from development, listing to sales, and maximize their market value. Product operation management includes planning and optimization of product market research, user demand analysis, marketing promotion, data analysis, after-sales service and other aspects. Effective product operation can help companies quickly respond to market demand, improve product competitiveness, optimize resource allocation, and increase product market share and profitability. Therefore, refined management of product operations has become one of the keys to the success of modern enterprises. Product operation management based on knowledge graphs improves the decision-making efficiency and accuracy of product operations by building an associated network of various data and knowledge related to products. Knowledge graphs can associate and integrate data from different sources such as product information, user needs, market dynamics, etc. to form a comprehensive and intuitive knowledge structure, thereby providing in-depth analysis and intelligent decision-making support for product operations. Through methods based on knowledge graphs, companies can more accurately understand market trends, user behaviors and product performance, and then optimize operational strategies and execution processes.

[0003] Existing product operation management technologies based on knowledge graphs mainly rely on knowledge graphs to build comprehensive data models and knowledge bases for each link in the product operation process. Specifically, these technologies deeply integrate and associate multiple data sources such as product design, production, sales, and customer feedback to create a multi-dimensional knowledge structure to support the management of each stage of the product life cycle. For example, knowledge graphs can help companies extract key insights from market research and user demand analysis, and compare this information with product characteristics and competitor status to provide rich data support for operational decisions. By semantically processing and intelligently reasoning information from different dimensions, knowledge graphs can automatically identify potential market opportunities, customer pain points, and operational bottlenecks, thereby optimizing operational activities such as product pricing, promotion strategies, and customer relationship management. In addition, knowledge graphs can also help companies quickly adjust their operational strategies and improve the adaptability and competitiveness of products in the market by monitoring market trends and user feedback in real time. Therefore, product operation management technologies based on knowledge graphs can not only improve operational efficiency, but also provide companies with accurate decision-making basis to help them maintain their leading position in a highly competitive market environment.

[0004] The prior art has the following deficiencies:

[0005] In a multi-category e-commerce platform, when a user browses a certain product (such as a smartwatch) and the product is not available for purchase due to being sold out or the end of the promotion, the system needs to recommend alternative products to meet the user's needs. However, in this case, the recommendation of alternative products often cannot accurately reflect the user's acceptance of multi-dimensional characteristics such as function, price, and brand. Since the existing product operation management technology based on knowledge graphs only relies on the basic attribute labels of the product (such as brand, price range) for simple matching, it fails to comprehensively consider the complex correlation between user needs and product characteristics through advanced mathematical modeling, especially lacks in-depth prediction of the user's acceptance of alternative products, resulting in deviations in the recommended alternative products in matching the user's actual needs. This lack of ability will lead to user dissatisfaction with the recommended content, thereby reducing the user's trust in the platform recommendation system and the user experience, which may lead to a decrease in purchase conversion rate, user loss, and waste of platform marketing resources.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a product operation management method and system based on knowledge graph to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a product operation management method based on knowledge graph, which specifically includes the following steps:

[0009] In a multi-category e-commerce platform, users’ purchasing behavior is monitored in real time. If the product that the user intends to purchase is not available, all alternative products associated with the product are determined;

[0010] Obtain comprehensive characteristic information of each alternative product in real time, analyze it after acquisition, and generate functional matching coefficient and user acceptance index of each alternative product;

[0011] Constructing a substitution evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generating a substitution evaluation coefficient for each substitute product, and analyzing it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and classifying each substitute product into a high-substitutability product, a potential substitute product, and a low-substitutability product according to the evaluation results;

[0012] According to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitutable products and low substitutable products;

[0013] Continuously monitor user behavioral feedback on recommended products, dynamically adjust alternative evaluation models and management measures based on real-time monitoring results, and store and analyze historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

[0014] Preferably, the comprehensive characteristic information of each substitute product is obtained in real time, and analyzed after being obtained to generate the function matching coefficient and user acceptance index of each substitute product respectively, which specifically includes the following steps:

[0015] Obtain comprehensive characteristic information of each substitute product in real time and perform preprocessing after acquisition;

[0016] Extracting market performance information and user behavior information from the pre-processed comprehensive characteristic information of each substitute product;

[0017] The market performance information and user behavior information in the extracted comprehensive characteristic information of each substitute product are analyzed to generate a functional matching coefficient and a user acceptance index for each substitute product.

[0018] Preferably, the logic for obtaining the functional matching coefficient of each alternative product is as follows:

[0019] Extract the market performance information from the pre-processed comprehensive characteristic information of each substitute product, including the price of the unavailable product and each substitute product, the historical sales volume of the unavailable product and each substitute product, and the inventory quantity of each substitute product when the product that the user intends to purchase is unavailable, and mark them as PM, PN respectively. i , SM, SN i and DN i , PM represents the price of the unavailable item when the item the user intends to purchase is unavailable, PN i represents the price of the i-th alternative product when the product that the user intends to purchase is not available, SM represents the historical sales volume of the unavailable product when the product that the user intends to purchase is not available, and SN i represents the historical sales volume of the ith alternative product when the product that the user intends to purchase is no longer available, DN i represents the inventory quantity of the i-th alternative product when the product that the user intends to purchase is not available, i = 1, 2, 3, ..., k, k is a positive integer;

[0020] Calculate the functional matching coefficient of each alternative product. The specific calculation formula is as follows:

[0021]

[0022] Where, FMC iis the functional matching coefficient of the ith alternative product.

[0023] Preferably, the logic for obtaining the user acceptance index of each alternative product is as follows:

[0024] Extract the user behavior information from the pre-processed comprehensive characteristic information of each alternative product, including the total number of clicks, total purchases, and total display times of each alternative product on the platform during the historical period, and mark them as CN i GN i and EN i , CN i represents the total number of clicks on the i-th alternative product by users in the historical period, GN i represents the total number of purchases of the i-th alternative product by the user in the historical period, EN i represents the total number of times the ith alternative product is displayed on the platform during the historical period, i = 1, 2, 3, ..., k, where k is a positive integer;

[0025] Calculate the user acceptance index of each alternative product. The specific calculation formula is as follows:

[0026]

[0027] Where UAI i is the user acceptance index of the i-th alternative product.

[0028] Preferably, constructing a substitution evaluation model for the function matching coefficients and user acceptance indexes of the generated substitute products to generate substitution evaluation coefficients of the substitute products specifically includes the following steps:

[0029] Collect the functional matching coefficients, user acceptance indexes and corresponding alternative evaluation coefficients of several alternative products generated in the past period of time, and calibrate them as and x represents the function matching coefficients, user acceptance indexes and corresponding substitution evaluation coefficients of several substitute products generated in the past period of time, x = 3, 4, 5, ..., d, d is a positive integer, and the data collected in the past period of time are formed into a historical data set;

[0030] The multivariate regression model is selected as an alternative evaluation model, and trained with historical data sets to determine the value of the regression coefficient according to the formula:

[0031]

[0032] Where β0, β1 and β2 are regression coefficients;

[0033] By minimizing the error between the predicted value and the actual value, the regression coefficient is optimized, and finally the values ​​of the regression coefficients β0, β1 and β2 are determined;

[0034] Using the finalized regression coefficient, the constructed alternative evaluation model is input with the functional matching coefficient FMC of each alternative product generated in real time. i and User Acceptance Index (UAI) i , generate the substitution evaluation coefficient SEC of each substitute product in real time i .

[0035] Preferably, the substitution evaluation coefficient SEC of each substitute commodity generated is i With the pre-set alternative evaluation coefficient threshold interval [SEC min , SEC max ] to compare, evaluate the degree of substitution of each substitute product for the non-purchasable product based on the comparison results, and classify each substitute product into high-substitutability product, potential substitute product and low-substitutability product based on the evaluation results. The specific comparison analysis and classification are as follows:

[0036] If the SEC i <SEC min , the degree of substitution of the substitute product for the non-purchasable product is low, and the substitute product is classified as a low-substitutability product;

[0037] If the SEC min ≤SEC i ≤SEC max , the degree of substitution of the substitute product for the non-purchasable product is moderate, and the substitute product is classified as a potential substitute product;

[0038] If the SEC i >SEC max , the degree of substitution of the substitute commodity for the non-purchasable commodity is high, and the substitute commodity is classified as a high-substitutability commodity.

[0039] Preferably, according to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitute products and low substitutable products, specifically:

[0040] For substitute products classified as low-substitutability products, the specific management measures are as follows: remove the product from the recommendation list to reduce its negative impact on user experience, and monitor the changes in the market performance information and user behavior information of the product at the same time. If its substitution evaluation coefficient increases, re-evaluate its substitution status;

[0041] For substitute products that are classified as potential substitute products, the specific management measures are as follows: the product is listed as the second-best recommendation option in the recommendation list, and detailed product descriptions, user reviews and promotional information are attached to increase users' awareness of its functions and value. At the same time, the market performance and user behavior data of the product are dynamically monitored, and its priority in the recommendation is regularly adjusted to optimize the recommendation effect;

[0042] For substitute products that are classified as highly substitutable products, the specific management measures taken are: give priority to the product as a recommendation option and attach reasons for the recommendation when recommending it to enhance user trust and acceptance. At the same time, its inventory changes and market performance are monitored in real time to ensure that the recommendation strategy is adjusted in time when the product inventory decreases and the status changes, so as to maintain continuous optimization of the user experience.

[0043] Preferably, a product operation management system based on knowledge graph includes a user behavior monitoring module, an alternative product information analysis module, an alternative evaluation modeling module, an alternative product management module and a dynamic feedback optimization module;

[0044] The user behavior monitoring module monitors the user's purchasing behavior in real time on a multi-category e-commerce platform. If the product that the user intends to purchase is not available, it determines all the alternative products associated with the product.

[0045] The substitute product information analysis module obtains the comprehensive characteristic information of each substitute product in real time, analyzes it after acquisition, and generates the function matching coefficient and user acceptance index of each substitute product respectively;

[0046] The substitutability evaluation modeling module constructs a substitutability evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generates a substitutability evaluation coefficient for each substitute product, and analyzes it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and divides each substitute product into high-substitutability products, potential substitute products, and low-substitutability products according to the evaluation results;

[0047] The substitute product management module implements different management measures for highly substitutable products, potential substitutable products and low substitutable products according to the classification results of each substitute product;

[0048] The dynamic feedback optimization module continuously monitors users' behavioral feedback on recommended products, dynamically adjusts alternative evaluation models and management measures based on real-time monitoring results, and stores and analyzes historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

[0049] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0050] 1. The present invention can quickly respond to user needs and accurately determine alternative products associated with non-purchasable products by monitoring the user's purchasing behavior and the status of non-purchasable products in real time. Combined with the calculation method of the function matching coefficient and the user acceptance index, a multi-dimensional comprehensive characteristic analysis is realized, overcoming the problem of inaccurate recommendation results caused by relying only on simple attribute labels in the prior art, and effectively improving the scientific nature of alternative product identification and the matching degree of user needs.

[0051] 2. The present invention constructs an alternative evaluation model and generates alternative evaluation coefficients, and uses a multivariate regression model to conduct a quantitative analysis of functional matching and user acceptance, thereby achieving an accurate assessment of the degree of substitution of alternative products. At the same time, the alternative product categories are divided according to the evaluation results, and differentiated management measures are taken, which not only optimizes the recommendation strategy, but also achieves efficient use of recommendation resources. Highly substitutable products are recommended first, potential substitute products enhance user awareness through additional information, and low-substitutability products are eliminated, thereby significantly improving user acceptance and trust in recommended content, improving user experience and promoting platform purchase conversion rates.

[0052] 3. The present invention realizes self-optimization of the system by continuously monitoring user behavioral feedback, dynamically adjusting alternative evaluation models and management logic, and storing and analyzing historical data. The combination of real-time feedback and long-term data ensures that the recommendation model and management logic can adapt to the dynamic changes of user needs and market environment, further enhancing the robustness of the system and the accuracy of the recommendation effect. This dynamic optimization mechanism not only improves the intelligence level of the recommendation system, but also greatly improves the operational efficiency and user satisfaction of the e-commerce platform, and has significant economic benefits and technology promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0054] Figure 1 This is a flow chart of a product operation management method and system based on knowledge graph of the present invention.

[0055] Figure 2 This is a module schematic diagram of a product operation management method and system based on knowledge graph in the present invention. DETAILED DESCRIPTION

[0056] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0057] The present invention provides Figure 1 The product operation management method based on knowledge graph shown in the figure specifically includes the following steps:

[0058] In a multi-category e-commerce platform, users’ purchasing behavior is monitored in real time. If the product that the user intends to purchase is not available, all alternative products associated with the product are determined;

[0059] In a multi-category e-commerce platform, real-time monitoring of user purchasing behavior can be achieved through an event-driven real-time capture mechanism. Specific methods include: embedding a user behavior tracking module in the front end of the platform, and using event listeners to capture user behaviors such as clicking the "Buy" button, selecting specifications, and submitting orders in real time, and marking these behaviors as purchase intentions. The captured behavior events will be pushed to the back-end stream processing system (such as Kafka or Flink) in real time to ensure that key information such as the timestamp of the operation, user ID, and product ID can be recorded and processed. In addition, the system will analyze the user's full path behavior from browsing to purchasing in combination with the operation log, and identify multi-step operations related to purchases. This method can capture the user's purchase intention with extremely low latency and provide real-time input for subsequent steps.

[0060] When it is detected that the product that the user intends to purchase is not available, the system quickly determines all associated alternative products through the knowledge graph. First, through the API interface of the product inventory management system and the promotion management system, the inventory status and promotion effectiveness of the product are queried in real time to determine whether the product is unavailable for purchase due to being sold out or the end of the activity. Once it is determined that the product is not available for purchase, the system will use the ID of the product as input to retrieve product nodes in the knowledge graph that have functional associations, brand associations, or user behavior associations with it. The knowledge graph generates a list of candidate alternative products by establishing attribute relationships between products (such as functional similarity and price range proximity). Subsequently, the system conducts a preliminary screening of candidate products based on the strength of association to ensure that the set of alternative products has high functional matching and recommendation value.

[0061] This design is to accurately solve the user's replacement needs in the scenario of unavailable goods, and to address the problem of "inaccurate alternative evaluation" in existing technologies. Real-time monitoring of user purchasing behavior ensures that the system can quickly perceive the unavailable status of goods and avoid user loss due to response delays. By retrieving alternative goods in the knowledge graph, the system can not only quickly locate candidate goods, but also lay a data foundation for subsequent alternative evaluations and recommendations. This method ensures the speed and accuracy of alternative product recommendations, improves user experience, and ultimately achieves an increase in the platform's purchase conversion rate and efficient use of resources.

[0062] Obtain comprehensive characteristic information of each alternative product in real time, analyze it after acquisition, and generate functional matching coefficient and user acceptance index of each alternative product;

[0063] In this embodiment, the comprehensive characteristic information of each substitute product is obtained in real time, and analyzed after being obtained to generate the function matching coefficient and user acceptance index of each substitute product respectively, which specifically includes the following steps:

[0064] Obtain comprehensive characteristic information of each substitute product in real time and perform preprocessing after acquisition;

[0065] The real-time acquisition of comprehensive characteristic information of substitute goods can be achieved through the real-time data collection system deployed on the e-commerce platform. Specific methods include: the system accesses the API interface of the product database, user behavior log system and inventory management system, and obtains the market performance information (such as price, sales volume, inventory) and user behavior information (such as number of clicks, number of purchases, number of impressions) of the product in a combination of timed polling and event-driven. The event-driven mechanism can monitor product status change events (such as price adjustments, inventory changes) and trigger data collection in real time to ensure the timeliness of information. At the same time, the system analyzes the data generated by the interaction between users and products (such as click logs, order records, etc.) in real time through the log parsing module, and stores the acquired data information in the data cache pool for subsequent processing. This method uses a highly automated software process to achieve real-time data acquisition and update, adapting to the needs of the dynamic environment of the e-commerce platform.

[0066] Data preprocessing is a key step to improve data quality, eliminate noise and unify the format, which directly affects the accuracy and efficiency of subsequent parameter calculation. The necessity of preprocessing is reflected in the following: the original data may be redundant, inconsistent or in a non-uniform format. For example, the price information of some commodities is denominated in US dollars and some in RMB, or the number of clicks may contain erroneous data (such as duplicate records). Through preprocessing, these problems can be processed into a standardized and defect-free data format, thereby providing high-quality input for analysis and parameter calculation. Specific preprocessing operations include: data cleaning (deleting duplicate or invalid records, such as outliers with negative inventory), data conversion (such as converting different currency units to the target currency unit), and data completion (using interpolation or mean filling for some missing values). Preprocessing operations are implemented through the automated process of the data pipeline. For example, the preprocessing module is triggered when data is collected in real time, and operations such as cleaning, conversion, and completion are embedded in the stream processing nodes of the pipeline to ensure that the data has high consistency and integrity before entering the analysis process.

[0067] Extracting market performance information and user behavior information from the pre-processed comprehensive characteristic information of each substitute product;

[0068] The market performance information and user behavior information in the extracted comprehensive characteristic information of each substitute product are analyzed to generate a functional matching coefficient and a user acceptance index for each substitute product.

[0069] In this embodiment, the logic for obtaining the function matching coefficient of each substitute product is as follows:

[0070] Extract the market performance information from the pre-processed comprehensive characteristic information of each substitute product, including the price of the unavailable product and each substitute product, the historical sales volume of the unavailable product and each substitute product, and the inventory quantity of each substitute product when the product that the user intends to purchase is unavailable, and mark them as PM, PN respectively. i , SM, SN i and DN i , PM represents the price of the unavailable item when the item the user intends to purchase is unavailable, PN i represents the price of the i-th alternative product when the product that the user intends to purchase is not available, SM represents the historical sales volume of the unavailable product when the product that the user intends to purchase is not available, and SN i represents the historical sales volume of the ith alternative product when the product that the user intends to purchase is no longer available, DN i represents the inventory quantity of the i-th alternative product when the product that the user intends to purchase is not available, i = 1, 2, 3, ..., k, k is a positive integer;

[0071] Extracting market performance information can be achieved by accessing the database and real-time data stream processing system of the e-commerce platform. Specifically, the system first calls the API interface of the product management system, takes the ID of the unavailable product as input, and retrieves the corresponding list of alternative products from the knowledge graph or product association relationship. Subsequently, by connecting to the inventory management module, the inventory quantity of each alternative product is queried in real time, and the results are stored in a temporary cache for subsequent use. At the same time, the system connects with the sales data system to extract the historical sales data of unavailable products and alternative products, which are usually automatically generated from sales records and updated regularly. Finally, the current market prices of unavailable products and alternative products are obtained through the price monitoring system, and the price data is synchronized in real time by the product management module to reflect the latest pricing of the product. The entire extraction process relies on an event-driven mechanism to ensure that data collection can be completed in a timely manner when user operations generate demand, and optimize data extraction efficiency through caching and multi-threaded processing.

[0072] The prices, historical sales, and inventory quantities of non-purchased goods and alternative goods can be automatically obtained through standardized interfaces. The prices of non-purchased goods and alternative goods come from the product management system and are obtained in real time by calling the price query interface. The price data is stored in the pricing database of the e-commerce platform, reflecting the current sales price of the goods; the historical sales data comes from the sales data system, which generates sales statistics based on the user's order records to ensure that the data is accurate and reliable; the inventory quantity is queried in real time through the inventory management module, which is connected to the inventory log of the warehousing system to dynamically reflect the entry, exit, and inventory changes of the goods. These data are obtained through a unified interface, and the data processing process does not rely on human intervention, which can achieve efficient and automated collection, ensuring the timeliness and integrity of market performance information.

[0073] Calculate the functional matching coefficient of each alternative product. The specific calculation formula is as follows:

[0074]

[0075] Where, FMC i is the functional matching coefficient of the ith alternative product.

[0076] The method of calculating the functional matching coefficient is to comprehensively measure the market performance matching degree of each substitute product with the non-purchased product in terms of price, sales volume and inventory, so as to scientifically quantify the substitutability of the substitute product. Calculating the relative difference in prices can eliminate the dimensionality problem caused by the difference in commodity price ranges and quantify the difference as a positive value, reflecting the adaptability of alternative commodities in terms of economic attributes; It is used to measure the relative difference of historical sales. Through normalization, sales data of different orders of magnitude are standardized for comparison, so as to evaluate the market popularity of alternative products. i ) By taking the logarithm of the inventory quantity, the impact of the maximum value on the result is compressed, and at the same time, the goods with insufficient inventory are given a lower weight to ensure that the recommended goods have good supply capacity. Finally, the three results are summed up and averaged, so that the functional matching coefficient is evenly distributed among various dimensions, forming a comprehensive alternative evaluation index that can adapt to a variety of product categories and provide stable and scientific results for alternative evaluation.

[0077] Functional matching coefficient FMC of the i-th substitute product i It directly reflects the degree of market performance matching between the commodity and the non-purchaseable commodity in terms of price, sales volume and inventory, and its size is positively correlated with the degree of substitution of the substitute commodity for the non-purchaseable commodity. The larger the functional matching coefficient, the closer the substitute commodity is to the non-purchaseable commodity in price and can meet the user's expectations of economic attributes; the closer the historical sales volume is, the more popular it is in the market and the more recognized it is by users; at the same time, its inventory is sufficient to ensure supply capacity and reduce the risk of users being unable to obtain the commodity. Therefore, the higher the functional matching coefficient of a substitute commodity, the stronger its ability to meet user needs and replace non-purchaseable commodities, and it is a more ideal alternative option; conversely, commodities with lower functional matching coefficients differ greatly from non-purchaseable commodities in key dimensions of market performance, and their substitutability is correspondingly weaker, making them unsuitable as priority recommended commodities.

[0078] In this embodiment, the logic for obtaining the user acceptance index of each alternative product is as follows:

[0079] Extract the user behavior information from the pre-processed comprehensive characteristic information of each alternative product, including the total number of clicks, total purchases, and total display times of each alternative product on the platform during the historical period, and mark them as CN i GN i and EN i , CN i represents the total number of clicks on the i-th alternative product by users in the historical period, GN i represents the total number of purchases of the i-th alternative product by the user in the historical period, EN i represents the total number of times the ith alternative product is displayed on the platform during the historical period, i = 1, 2, 3, ..., k, where k is a positive integer;

[0080] Extracting user behavior information can be achieved through the log analysis system and user behavior tracking module of the e-commerce platform. Specifically, the system connects the front-end event tracking module and the back-end user behavior log database. The front-end module records the user's clicks, purchases, browsing and other behavioral events on the platform in real time, including event timestamps, user IDs, product IDs and other information, and sends these event data to the log analysis system. The back-end log analysis system extracts behavioral data related to alternative products from the log through streaming or batch tasks, and summarizes them according to the set historical time period, such as counting the total number of clicks, total purchases and total impressions of each alternative product by users, and generates a cleaned data table stored in the cache for subsequent calculation of the user acceptance index. The extraction process relies on the event log recording and real-time processing capabilities of the e-commerce platform to ensure the accuracy and timeliness of data extraction.

[0081] These three types of data are obtained in different ways and have clear sources. The total number of clicks is obtained through the front-end event tracking module. When a user clicks on a substitute product, the event tracking module records the click behavior in the user behavior log, and the back-end performs statistics by filtering the product ID and event type; the total number of purchases comes from the order management system. When a user completes the purchase of a product, the system records the product ID and quantity generated by the order. The log analysis system filters the orders related to substitute products from the order records for statistics; the total number of impressions is obtained through the advertising delivery system and page browsing logs. The advertising delivery system records the exposure log of each product display, while the page browsing log records the number of times the user loads the product details page. The log analysis system combines these two types of data to count the number of impressions of each substitute product. These data are automatically generated through log analysis and statistical calculations to ensure the authenticity and real-time nature of the data.

[0082] Calculate the user acceptance index of each alternative product. The specific calculation formula is as follows:

[0083]

[0084] Where UAI i is the user acceptance index of the i-th alternative product.

[0085] The calculation of the user acceptance index is to comprehensively measure the user's preference for each alternative product and the conversion effect in the historical period. Each operation in the formula is designed for specific behavioral data characteristics to ensure the scientificity and accuracy of the evaluation results. i Take the logarithm ln(1+CN i ) is to reduce the excessive impact of extremely high click volume on the calculation results, and at the same time give a more obvious distinction to low-click volume products to avoid numerical imbalance; secondly, the purchase frequency GNi Take the square root The purpose is to amplify the weight of purchase behavior in the evaluation, because purchase behavior can better reflect the user's true preference and acceptance than click behavior. The introduction of square root can balance the differences between the number of purchases while maintaining a moderate numerical range. Finally, The normalized ratio of clicks and purchases to impressions directly quantifies the conversion rate of alternative products and shows their actual efficiency in being accepted by users. The user acceptance index generated by weighted average of these three indicators can not only comprehensively evaluate the comprehensive performance of user behavior, but also enhance the robustness and accuracy of the model through nonlinear transformations in the calculation steps, providing reliable parameter support for alternative evaluation.

[0086] User acceptance index UAI of the i-th alternative product i It directly reflects the comprehensive degree of acceptance and preference of the product by users in the historical period, and its size is positively correlated with "evaluating the degree of substitution of each alternative product for the non-purchaseable product". The larger the user acceptance index, the better the performance of the alternative product in key behavioral data such as clicks, purchases and displays, and can attract the attention of users and promote actual purchasing behavior, which means that it is more popular among users. Therefore, an alternative product with a higher user acceptance index is more likely to meet the actual needs of users and gain higher acceptance, thus gaining an advantage in the evaluation of the substitutability of non-purchaseable products; conversely, products with a lower user acceptance index are weaker in user behavior performance, lack of attractiveness and conversion ability, and naturally poorer in substitutability, and are not suitable as priority recommended alternative products.

[0087] Constructing a substitution evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generating a substitution evaluation coefficient for each substitute product, and analyzing it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and classifying each substitute product into a high-substitutability product, a potential substitute product, and a low-substitutability product according to the evaluation results;

[0088] In this embodiment, a substitution evaluation model is constructed based on the function matching coefficient and user acceptance index of each generated substitute product to generate a substitution evaluation coefficient of each substitute product, specifically including the following steps:

[0089] Collect the functional matching coefficients, user acceptance indexes and corresponding alternative evaluation coefficients of several alternative products generated in the past period of time, and calibrate them as and x represents the function matching coefficients, user acceptance indexes and corresponding substitution evaluation coefficients of several substitute products generated in the past period of time, x = 3, 4, 5, ..., d, d is a positive integer, and the data collected in the past period of time are formed into a historical data set;

[0090] The real-time data recording system and historical data management module of the e-commerce platform can be used to collect the functional matching coefficients, user acceptance indexes, and corresponding alternative evaluation coefficients of each alternative product generated in the past period of time. Specifically, in the process of generating the functional matching coefficients and user acceptance indexes in real time, the system stores the results of each calculation together with the timestamp and product ID in a database with a time index. At the same time, when generating the alternative evaluation coefficients, the system stores the corresponding functional matching coefficients, user acceptance indexes, and alternative evaluation coefficients as complete records in the historical data set. In order to improve query efficiency, the data will be automatically partitioned and managed according to the time period and product ID when it is stored, ensuring that when it is necessary to trace back the data of a period of time in the past, the target data can be quickly extracted by calling the time period range query interface. In this automated way, the system can continuously accumulate and manage the functional matching coefficients, user acceptance indexes, and alternative evaluation coefficients, ensure the integrity and availability of historical data, and provide high-quality input for the training of alternative evaluation models.

[0091] It is reasonable and necessary to limit the value of x to a positive integer greater than or equal to 3, because three regression coefficients β0, β1 and β2 need to be determined in the target equation, and the multivariate regression model solves the system of equations to obtain the optimal values ​​of these coefficients. Only when the historical data contains at least three data points (i.e., x ≥ 3) can three independent equations be formed to form a solvable system of equations, thereby accurately calculating the values ​​of the three regression coefficients. If x < 3, the data points are not enough to support the solution of the system of equations, and the model cannot be effectively trained. In addition, increasing the range of x values ​​can also improve the robustness and fitting ability of the model, reduce errors and optimize regression results through more historical data points. This value limitation not only ensures the basic solvability of the model, but also provides data support for subsequent error optimization and the accuracy of regression coefficients, so it is scientific and reasonable.

[0092] The multivariate regression model is selected as an alternative evaluation model, and trained with historical data sets to determine the value of the regression coefficient according to the formula:

[0093]

[0094] Where β0, β1 and β2 are regression coefficients;

[0095] The multiple regression model is a statistical modeling method that predicts or explains the changes in the dependent variable through the linear combination of multiple independent variables. The core of the model is to construct a linear equation that combines the contribution weights of each variable (i.e., regression coefficients) with their numerical values ​​to obtain an estimated value of the dependent variable. In the alternative evaluation, the multiple regression model is selected as the alternative evaluation model because the functional matching coefficient and User Acceptance Index are two key independent variables that directly affect the alternative evaluation coefficient The size of this dependent variable. By using historical data sets to train the model, the values ​​of the regression coefficients can be optimized and determined from actual data to ensure that the model has good predictive ability and practical applicability. The three regression coefficients in the model have clear meanings: β0 is a constant term, which represents the baseline value of the alternative evaluation coefficient when the function matching coefficient and the user acceptance index are both zero; β1 is the weight of the function matching coefficient, which reflects the importance of function matching in alternative evaluation and its influence on the final evaluation coefficient; β2 is the weight of the user acceptance index, which indicates the effect of user behavior preferences on alternative evaluation results. By minimizing the error between the predicted value and the actual value and optimizing the values ​​of these regression coefficients, the combined impact of function matching and user acceptance on alternative evaluation can be captured to the maximum extent, ensuring the model's ability to accurately evaluate real-time data.

[0096] By minimizing the error between the predicted value and the actual value, the regression coefficient is optimized, and finally the values ​​of the regression coefficients β0, β1 and β2 are determined;

[0097] The purpose of optimizing the regression coefficient by minimizing the error between the predicted value and the actual value is to ensure that the multivariate regression model can accurately capture the actual impact of the functional matching coefficient and the user acceptance index on the alternative evaluation coefficient. The core of this process is to evaluate the difference between the model prediction results and the actual results of the historical data, and to reduce this difference by repeatedly adjusting the regression coefficient so that the model can better fit the actual data. Specifically, an indicator (such as an error metric) is constructed to measure the gap between the predicted value and the actual value, and then an iterative optimization algorithm (such as the gradient descent method) is used to gradually adjust the regression coefficient to minimize the error. Through such an optimization process, the final regression coefficient can fully reflect the contribution of the functional matching coefficient and the user acceptance index in the historical data to the alternative evaluation coefficient, making the model more predictive and more reliably applicable to real-time data analysis.

[0098] Using the finalized regression coefficient, the constructed alternative evaluation model is input with the functional matching coefficient FMC of each alternative product generated in real time. i and User Acceptance Index (UAI) i , generate the substitution evaluation coefficient SEC of each substitute product in real time i.

[0099] In this embodiment, the generated substitution evaluation coefficient SEC of each substitution commodity is i With the pre-set alternative evaluation coefficient threshold interval [SEC min , SEC max ] to compare, evaluate the degree of substitution of each substitute product for the non-purchasable product based on the comparison results, and classify each substitute product into high-substitutability product, potential substitute product and low-substitutability product based on the evaluation results. The specific comparison analysis and classification are as follows:

[0100] If the SEC i <SEC min , the degree of substitution of the substitute product for the non-purchasable product is low, and the substitute product is classified as a low-substitutability product;

[0101] This situation indicates that the product is not an effective substitute for non-purchaseable products in terms of functional matching and user acceptance. This means that it is quite different from non-purchaseable products in terms of market performance information such as price, sales volume and inventory, and it also lacks sufficient support from users for its behavior, such as low click-through rate and purchase rate. Even if such products are recommended to users, they are unlikely to meet their actual needs, which may reduce users' satisfaction and trust in the recommendation system and waste recommendation resources. Therefore, such products are usually classified as low-substitutability products, with a lower priority in recommendations, or even removed from the recommendation list to avoid negative impact on user experience.

[0102] If the SEC min ≤SEC i ≤SEC max , the degree of substitution of the substitute product for the non-purchasable product is moderate, and the substitute product is classified as a potential substitute product;

[0103] This situation indicates that the performance of this product in terms of functional matching and user acceptance is closer to that of non-purchaseable products, but there is still room for improvement. This means that such products meet user needs in some market performance indicators (such as price, sales volume or inventory), but are still insufficient in other aspects (such as user behavior indicators). Such products have a certain possibility of meeting user needs, but need to be supplemented with more information guidance, such as highlighting the advantages and characteristics of the product or user reviews, in order to enhance user trust and willingness to buy. Therefore, such products will be classified as potential alternative products, which can be displayed as alternative options in recommendations, and user acceptance can be improved by providing additional information.

[0104] If the SEC i >SEC max, the degree of substitution of the substitute commodity for the non-purchasable commodity is high, and the substitute commodity is classified as a high-substitutability commodity.

[0105] This indicates that the product has excellent performance in terms of functional matching and user acceptance, and can well meet users' demand for alternatives to unavailable products. This means that such products are highly consistent with unavailable products in terms of market performance information such as price, sales volume, and inventory, and user behavior indicators (such as click-through rate and purchase rate) are also outstanding, indicating that users have a high degree of acceptance and preference for the product. Such products can effectively meet user needs and are the priority recommendation objects of the recommendation system. They can be directly displayed to users with clear reasons for recommendation (such as "functional matching and high cost performance"), thereby improving user experience, conversion rate, and platform revenue.

[0106] The preset threshold interval of the alternative evaluation coefficient can be determined by analyzing the distribution of the alternative evaluation coefficient in the historical data. The specific implementation method is to use data statistics and analysis software to extract the distribution characteristics of the historical data set. The system first collects the values ​​of all alternative evaluation coefficients from the historical data set, and analyzes their distribution according to quantiles or other statistical methods, such as using the common 5% and 95% quantiles to define the minimum and maximum values ​​of the interval respectively, thereby covering the range of most evaluation coefficients. If it is necessary to more accurately reflect the characteristics of different commodity categories, the alternative commodities can be grouped by category, and the distribution of the alternative evaluation coefficients of each group can be counted separately. The data area is divided using a cluster analysis algorithm or a kernel density estimation method to automatically extract the threshold interval suitable for the category. In addition, the system can regularly update the threshold interval according to the new historical data through a real-time dynamic adjustment method to ensure that the interval range can reflect the changes in the current market and user behavior. This automated threshold setting method based on historical data and distribution characteristics is not only scientific and reasonable, but also adaptable to changing market needs.

[0107] According to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitutable products and low substitutable products;

[0108] In this embodiment, according to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitute products and low substitutable products, specifically:

[0109] For substitute products classified as low-substitutability products, the specific management measures are as follows: remove the product from the recommendation list to reduce its negative impact on user experience, and monitor the changes in the market performance information and user behavior information of the product at the same time. If its substitution evaluation coefficient increases, re-evaluate its substitution status;

[0110] For low-substitutability products, management measures can be implemented through the priority management module and data monitoring module of the recommendation system. Specifically, the system first marks low-substitutability products as "low priority" based on the determination result of the substitution evaluation coefficient, and removes them from the recommendation list, or adjusts their sorting position to the lowest to reduce interference with user recommendations. The data monitoring module continuously collects market performance information (such as sales and inventory) and user behavior information (such as click and purchase data) of these products, and monitors the change range of the substitution evaluation coefficient by comparing historical data. When the system detects that the substitution evaluation coefficient has increased significantly and reaches the trigger condition for re-evaluation, it triggers the re-evaluation process, recalculates its substitution status and adjusts the management strategy. This is because low-substitutability products may have a negative impact on user experience, and by removing or lowering the priority, it can avoid wasting recommendation resources. At the same time, the dynamic monitoring and re-evaluation mechanism ensures that the system is adaptable to changes in product status.

[0111] For substitute products that are classified as potential substitute products, the specific management measures are as follows: the product is listed as the second-best recommendation option in the recommendation list, and detailed product descriptions, user reviews and promotional information are attached to increase users' awareness of its functions and value. At the same time, the market performance and user behavior data of the product are dynamically monitored, and its priority in the recommendation is regularly adjusted to optimize the recommendation effect;

[0112] For potential substitute products, management measures can be implemented through the suboptimal recommendation strategy module and additional information management module of the recommendation system. Specifically, the system marks the potential substitute products as "suboptimal recommendation" and assigns them a medium priority in the recommendation list. At the same time, the additional information management module adds detailed descriptions, user reviews, and promotional information (such as discounts and free shipping) to the products. These additional information are obtained by calling the product database, evaluation data system, and promotion management system, and are presented in the recommendation interface in the form of dynamic data binding. The system also regularly evaluates the substitution evaluation coefficient and dynamically adjusts the recommendation priority by monitoring the click-through rate, purchase rate, and inventory changes of the product in real time. The reason for this is that there is still a certain gap in the degree of substitution of potential substitute products, but additional information can enhance user cognition and trust, and the dynamic adjustment mechanism ensures that the recommendation strategy can be continuously optimized with changes in the market and user behavior.

[0113] For substitute products that are classified as highly substitutable products, the specific management measures taken are: give priority to the product as a recommendation option and attach reasons for the recommendation when recommending it to enhance user trust and acceptance. At the same time, its inventory changes and market performance are monitored in real time to ensure that the recommendation strategy is adjusted in time when the product inventory decreases and the status changes, so as to maintain continuous optimization of the user experience.

[0114] For highly substitutable products, management measures can be implemented through the priority recommendation strategy module and real-time monitoring module of the recommendation system. Specifically, the system marks highly substitutable products as "priority recommendation" and arranges them at the top of the recommendation list, while attaching clear reasons for recommendation (such as "high functional matching" and "favorable price"). The recommendation reason is generated by analyzing the functional matching coefficient and user acceptance index of the product, and extracting specific advantage information from the product database for dynamic presentation. The real-time monitoring module will continue to track the inventory, sales volume and user behavior changes of highly substitutable products. If a significant reduction in inventory or a change in status (such as a price adjustment or the end of a promotion) is detected, the recalculation process will be immediately triggered to adjust the recommendation strategy. This is done because highly substitutable products can best meet user needs, and their priority recommendation can improve user experience and purchase conversion rate, while the real-time monitoring mechanism ensures that the recommendation strategy can be adjusted in time when the product status changes to avoid user dissatisfaction due to invalid recommendations.

[0115] Continuously monitor user behavioral feedback on recommended products, dynamically adjust alternative evaluation models and management measures based on real-time monitoring results, and store and analyze historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

[0116] The user's behavioral feedback on recommended products can be monitored and processed through the real-time behavior tracking module and log analysis system. Specifically, when the user interacts with the recommended products, the system captures the user's click, purchase, favorite or skip operations in real time through the front-end behavior tracking module, and records the operation timestamp, product ID, user ID and other key information. These behavioral events will be pushed to the log analysis system in real time for processing. The system calculates the user's feedback intensity on the recommended products (such as the growth rate of clicks and purchases) based on different event types. The feedback intensity data will be updated in real time to the alternative evaluation model, and the model will be dynamically fine-tuned to enable the model to adapt to user behavior changes more quickly. This method can ensure that the recommendation system always adjusts the evaluation and recommendation strategies based on the latest user feedback, thereby improving the accuracy of the recommendation results and user satisfaction.

[0117] The storage and analysis of historical data can be achieved through the distributed database and big data analysis system of the e-commerce platform. Specifically, the system records the relevant data of user behavior feedback, product market performance and recommendation strategy adjustment into a database with a timestamp, and partitions and stores them according to product category or recommendation type to facilitate subsequent queries. Through the integrated big data analysis system, historical data is batch processed regularly to analyze long-term user behavior trends and recommendation strategy effects. For example, the system can calculate key indicators such as the average click-through rate and purchase conversion rate of different product categories to evaluate the overall performance of the current recommendation strategy. These analysis results will serve as the basis for alternative evaluation models and management logic optimization, providing specific directions for model improvement. Doing so can combine short-term real-time feedback with long-term historical data, helping the system build a more comprehensive recommendation optimization system.

[0118] Continuously monitoring user behavior feedback and dynamically adjusting models and management logic is to enable the recommendation system to always respond quickly to user needs and market changes, thereby improving the accuracy of recommendations and user experience. Real-time behavioral feedback monitoring ensures that the system can capture users' immediate reactions to current recommendation strategies, dynamically adjust alternative evaluation models and recommendation priorities, and ensure that recommendation results match user interests. At the same time, historical data storage and analysis provide the system with a wider range and longer-term reference basis, and further optimize model parameters and management logic by summarizing user behavior patterns and the effects of recommendation strategies. This combination of real-time and historical methods can not only ensure rapid response in the short term, but also improve long-term recommendation efficiency and stability, helping the platform to remain competitive in the ever-changing user needs and market environment.

[0119] like Figure 2 A product operation management system based on knowledge graph is shown, including a user behavior monitoring module, an alternative product information analysis module, an alternative evaluation modeling module, an alternative product management module and a dynamic feedback optimization module;

[0120] The user behavior monitoring module monitors the user's purchasing behavior in real time on a multi-category e-commerce platform. If the product that the user intends to purchase is not available, it determines all the alternative products associated with the product.

[0121] The substitute product information analysis module obtains the comprehensive characteristic information of each substitute product in real time, analyzes it after acquisition, and generates the function matching coefficient and user acceptance index of each substitute product respectively;

[0122] The substitutability evaluation modeling module constructs a substitutability evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generates a substitutability evaluation coefficient for each substitute product, and analyzes it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and divides each substitute product into high-substitutability products, potential substitute products, and low-substitutability products according to the evaluation results;

[0123] The substitute product management module implements different management measures for highly substitutable products, potential substitutable products and low substitutable products according to the classification results of each substitute product;

[0124] The dynamic feedback optimization module continuously monitors users' behavioral feedback on recommended products, dynamically adjusts alternative evaluation models and management measures based on real-time monitoring results, and stores and analyzes historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

[0125] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0126] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0127] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0128] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0129] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0130] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0131] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0132] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A product operation management method based on knowledge graph, characterized in that: The specific steps include: In a multi-category e-commerce platform, users’ purchasing behavior is monitored in real time. If the product that the user intends to purchase is not available, all alternative products associated with the product are determined; Obtain comprehensive characteristic information of each alternative product in real time, analyze it after acquisition, and generate functional matching coefficient and user acceptance index of each alternative product; Constructing a substitution evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generating a substitution evaluation coefficient for each substitute product, and analyzing it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and classifying each substitute product into a high-substitutability product, a potential substitute product, and a low-substitutability product according to the evaluation results; According to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitutable products and low substitutable products; Continuously monitor user behavioral feedback on recommended products, dynamically adjust alternative evaluation models and management measures based on real-time monitoring results, and store and analyze historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

2. According to the product operation management method based on knowledge graph in claim 1, it is characterized in that: The comprehensive characteristic information of each alternative product is obtained in real time, and analyzed after acquisition to generate the functional matching coefficient and user acceptance index of each alternative product respectively, which specifically includes the following steps: Obtain comprehensive characteristic information of each substitute product in real time and perform preprocessing after acquisition; Extracting market performance information and user behavior information from the pre-processed comprehensive characteristic information of each substitute product; The market performance information and user behavior information in the extracted comprehensive characteristic information of each substitute product are analyzed to generate a functional matching coefficient and a user acceptance index for each substitute product.

3. A product operation management method based on knowledge graph according to claim 2, characterized in that: The logic for obtaining the functional matching coefficient of each alternative product is as follows: Extract the market performance information from the pre-processed comprehensive characteristic information of each substitute product, including the price of the unavailable product and each substitute product, the historical sales volume of the unavailable product and each substitute product, and the inventory quantity of each substitute product when the product that the user intends to purchase is unavailable, and mark them as PM, PN respectively. i , SM, SN i and DN i , PM represents the price of the unavailable item when the item the user intends to purchase is unavailable, PN i represents the price of the i-th alternative product when the product that the user intends to purchase is not available, SM represents the historical sales volume of the unavailable product when the product that the user intends to purchase is not available, and SN i represents the historical sales volume of the ith alternative product when the product that the user intends to purchase is no longer available, DN i represents the inventory quantity of the i-th alternative product when the product that the user intends to purchase is not available, i = 1, 2, 3, ..., k, k is a positive integer; Calculate the functional matching coefficient of each alternative product. The specific calculation formula is as follows: Where, FMC i is the functional matching coefficient of the ith alternative product.

4. A product operation management method based on knowledge graph according to claim 3, characterized in that: The logic for obtaining the user acceptance index of each alternative product is as follows: Extract the user behavior information from the pre-processed comprehensive characteristic information of each alternative product, including the total number of clicks, total purchases, and total display times of each alternative product on the platform by users in the historical period, and mark them as CN i GN i and EN i , CN i represents the total number of clicks on the i-th alternative product by users in the historical period, GN i represents the total number of purchases of the i-th alternative product by the user in the historical period, EN i represents the total number of times the ith alternative product is displayed on the platform during the historical period, i = 1, 2, 3, ..., k, where k is a positive integer; Calculate the user acceptance index of each alternative product. The specific calculation formula is as follows: Where UAI i is the user acceptance index of the i-th alternative product.

5. The product operation management method based on knowledge graph according to claim 4 is characterized in that: A substitution evaluation model is constructed based on the function matching coefficient and user acceptance index of each generated substitute product to generate a substitution evaluation coefficient of each substitute product, specifically including the following steps: Collect the functional matching coefficients, user acceptance indexes and corresponding alternative evaluation coefficients of several alternative products generated in the past period of time, and calibrate them as and x represents the function matching coefficients, user acceptance indexes and corresponding substitution evaluation coefficients of several substitute products generated in the past period of time, x = 3, 4, 5, ..., d, d is a positive integer, and the data collected in the past period of time are formed into a historical data set; The multivariate regression model is selected as an alternative evaluation model, and trained with historical data sets to determine the value of the regression coefficient according to the formula: Where β0, β1 and β2 are regression coefficients; By minimizing the error between the predicted value and the actual value, the regression coefficient is optimized, and finally the values ​​of the regression coefficients β0, β1 and β2 are determined; Using the finalized regression coefficient, the constructed alternative evaluation model is input with the functional matching coefficient FMC of each alternative product generated in real time. i and User Acceptance Index (UAI) i , generate the substitution evaluation coefficient SEC of each substitute product in real time i .

6. A product operation management method based on knowledge graph according to claim 5, characterized in that: The substitution evaluation coefficient SEC of each alternative commodity generated i With the pre-set alternative evaluation coefficient threshold interval [SEC min , SEC max ] to compare, evaluate the degree of substitution of each substitute product for the non-purchasable product based on the comparison results, and classify each substitute product into high-substitutability product, potential substitute product and low-substitutability product based on the evaluation results. The specific comparison analysis and classification are as follows: If the SEC i <SEC min , the degree of substitution of the substitute product for the non-purchasable product is low, and the substitute product is classified as a low-substitutability product; If the SEC min ≤SEC i ≤SEC max , the degree of substitution of the substitute product for the non-purchasable product is moderate, and the substitute product is classified as a potential substitute product; If the SEC i >SEC max , the degree of substitution of the substitute commodity for the non-purchasable commodity is high, and the substitute commodity is classified as a high-substitutability commodity.

7. The product operation management method based on knowledge graph according to claim 6 is characterized in that: According to the classification results of each substitute product, different management measures are taken for highly substitutable products, potential substitutable products and low substitutable products, specifically: For substitute products classified as low-substitutability products, the specific management measures are as follows: remove the product from the recommendation list to reduce its negative impact on user experience, and monitor the changes in the market performance information and user behavior information of the product at the same time. If its substitution evaluation coefficient increases, re-evaluate its substitution status; For substitute products that are classified as potential substitute products, the specific management measures are as follows: the product is listed as the second-best recommendation option in the recommendation list, and detailed product descriptions, user reviews and promotional information are attached to increase users' awareness of its functions and value. At the same time, the market performance and user behavior data of the product are dynamically monitored, and its priority in the recommendation is regularly adjusted to optimize the recommendation effect; For substitute products that are classified as highly substitutable products, the specific management measures taken are: give priority to the product as a recommendation option and attach reasons for the recommendation when recommending it to enhance user trust and acceptance. At the same time, its inventory changes and market performance are monitored in real time to ensure that the recommendation strategy is adjusted in time when the product inventory decreases and the status changes, so as to maintain continuous optimization of the user experience.

8. A product operation management system based on knowledge graph, used to implement a product operation management method based on knowledge graph as described in any one of claims 1 to 7, characterized in that: It includes user behavior monitoring module, alternative product information analysis module, alternative evaluation modeling module, alternative product management module and dynamic feedback optimization module; The user behavior monitoring module monitors the user's purchasing behavior in real time on a multi-category e-commerce platform. If the product that the user intends to purchase is not available, it determines all the alternative products associated with the product. The substitute product information analysis module obtains the comprehensive characteristic information of each substitute product in real time, analyzes it after acquisition, and generates the function matching coefficient and user acceptance index of each substitute product respectively; The substitutability evaluation modeling module constructs a substitutability evaluation model for the functional matching coefficients and user acceptance indexes of each generated substitute product, generates a substitutability evaluation coefficient for each substitute product, and analyzes it after generation to evaluate the degree of substitution of each substitute product for the non-purchased product, and divides each substitute product into high-substitutability products, potential substitute products, and low-substitutability products according to the evaluation results; The substitute product management module implements different management measures for highly substitutable products, potential substitutable products and low substitutable products according to the classification results of each substitute product; The dynamic feedback optimization module continuously monitors users' behavioral feedback on recommended products, dynamically adjusts alternative evaluation models and management measures based on real-time monitoring results, and stores and analyzes historical data to optimize model parameters and management logic, thereby improving the system's recommendation accuracy and user experience.

Images