Intelligent electronic commerce online management system and method thereof

Abstract

The invention discloses an intelligent e-commerce online management system and method, and relates to the technical field of e-commerce systems, and the method comprises the steps: collecting the data of a user terminal, extracting the situation data, combining the situation data with the historical data of a user, and generating a comprehensive feature vector; preference commodities and logistics nodes related to user demands are screened out according to the comprehensive feature vectors, and a logistics twinborn model is constructed; generating a preliminary recommendation list based on the preferred commodities, inputting the preliminary recommendation list into a logistics twinborn model, and performing screening adjustment through inventory states, distribution paths and logistics load data; collecting inventory, sales volume and logistics state data of the logistics twinborn model, analyzing a commodity sales trend and a logistics load condition in combination with the optimization recommendation list, predicting a future inventory demand and a logistics load, and generating an adjustment strategy; collecting feedback data of the user to the adjustment strategy, and optimizing the logistics twinborn model; and combining an optimization result with an adjustment strategy, and dynamically adjusting a recommendation strategy, a replenishment plan and a logistics path.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce systems, and particularly to an intelligent e-commerce online management system and method thereof. Background Art

[0002] With the rapid development of e-commerce, online shopping platforms have played an important role in meeting consumers' needs for goods and services. Currently, recommendation algorithms based on user behavior data (such as collaborative filtering, content-based recommendation, and deep learning algorithms) have been widely used. By analyzing users' historical behavior data, interest preferences are mined to generate a list of recommended products. At the same time, logistics management technology has also been continuously upgraded, using means such as warehouse management, distribution route optimization, and dynamic inventory control to improve logistics efficiency. However, with the diversification of user needs and the increasing complexity of the logistics supply chain, traditional recommendation algorithms and logistics management systems have gradually shown problems of insufficient collaboration.

[0003] The existing technologies mainly have the following deficiencies: Recommendation algorithms usually only rely on users' historical behavior data and are difficult to capture users' immediate needs and dynamic preferences in different situations, resulting in poor personalization and real-time performance of recommendations. In addition, there is a lack of linkage between the recommendation system and the logistics system, and the real-time status of logistics resources (such as inventory, distribution route, and load) is not effectively utilized to dynamically optimize the recommendation results. This fragmented system architecture may lead to insufficient inventory of recommended products, high distribution route costs, or overloaded logistics nodes, directly affecting the user experience and the overall operation efficiency of the platform. Summary of the Invention

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

[0005] Therefore, the present invention provides an intelligent e-commerce online management method to solve the problem that recommendation algorithms usually only rely on users' historical behavior data and are difficult to capture users' immediate needs and dynamic preferences in different situations, resulting in poor personalization and real-time performance of recommendations.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides an intelligent e-commerce online management method, which includes: Collect sensor data and operation behavior data of a user terminal, extract context data through a context awareness algorithm, and combine the context data with the user's historical data to obtain a comprehensive feature vector; According to the comprehensive feature vector, screen out preference products and logistics nodes related to the user's needs, and construct a logistics twin model by combining the preference products and the logistics nodes; Generate a preliminary recommendation list based on preferred products, input the preliminary recommendation list into the logistics twin model, and screen and adjust the recommendation list based on the inventory status, delivery route, and logistics load data of the logistics twin model to generate an optimized recommendation list; Collect the inventory, sales volume, and logistics status data of the logistics twin model, combine with the optimized recommendation list, analyze the commodity sales trend and logistics load situation, use prediction algorithms to predict future inventory demand and logistics load, and generate adjustment strategies; Collect the feedback data of users on the adjustment strategies, and optimize the logistics twin model through the feedback data; Combine the optimization results with the adjustment strategies to dynamically adjust the recommendation strategy, replenishment plan, and logistics route.

[0007] As an optimal solution of the intelligent e-commerce online management method described in the present invention, wherein: collecting the sensor data and operation behavior data of the user terminal, extracting context data through the context awareness algorithm, and combining the context data with the user historical data to obtain a comprehensive feature vector. The specific steps are as follows: Integrate the multi-modal data into a context feature vector through the context awareness algorithm by using the multi-modal sensor data and operation behavior data of the user terminal, combined with the timestamp information and environmental context, to obtain the context feature vector ; Collect the historical data of the user and perform frequency statistics on the browsing times, purchase times, and favorite times to generate a user historical feature vector ; The context feature vector is dynamically fused with the historical preference data of the user to generate a comprehensive feature vector , and the expression is: ; ; ; wherein, represents the comprehensive context feature vector calculated by weighted summation from the context feature vector , represents the weight of the th context feature, represents the th context feature vector at the time point , represents the historical preference score calculated by weighted summation from the user historical feature vector , represents the preference score, represents the th weight of the historical feature vector, Represents the preference score of the th historical feature vector, represents the weight factor of the comprehensive context feature vector, represents the weight factor of the historical feature vector; The comprehensive feature vector is directly mapped to the preference score of the specific commodity category through the regression model .

[0008] As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: the As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: according to the comprehensive feature vector, the preferred commodities and logistics nodes related to the user's needs are screened out, and a logistics twin model is constructed by combining the preferred commodities and logistics nodes. The specific steps are as follows: Set the preference threshold to , set the preference score of the th commodity to , screen out the commodity categories, and then perform a set operation to obtain the candidate commodity set ; Each commodity category corresponds to a logistics node. Set the logistics node set to , and the logistics node related to the commodity is ; Screen each commodity category in the candidate commodity category set ; When the inventory is insufficient, then eliminate this logistics node. When the current load of the logistics node exceeds the threshold, then eliminate this logistics node; The candidate commodity category set , the logistics node set and the user's delivery path set are combined to obtain the logistics twin model .

[0009] As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: generating a preliminary recommendation list based on the preferred commodities, inputting the preliminary recommendation list into the logistics twin model, and screening and adjusting the recommendation list through the inventory status, delivery path and logistics load data of the logistics twin model to generate an optimized recommendation list. The specific steps are as follows: Based on the candidate commodity category set , a preliminary recommendation list is generated by combining the sales volume factor, evaluation factor and price factor. The expression is: ; Among them, represents the commodity The comprehensive score in the preliminary recommendation list, represents the historical sales volume of the product , represents the maximum historical sales volume of all products, represents the evaluation score of the user for the product , represents the maximum evaluation score of all products, represents the average value of the preference scores of all products, represents the maximum preference score of all product categories, and are the weight factors of the preference score, historical sales volume, evaluation score, and price deviation respectively; Input the preliminary recommendation list into the logistics twin model to screen and adjust the preliminary recommendation list; Eliminate products with insufficient inventory; For the remaining products, calculate the distribution path cost from the logistics node to the user, and obtain the total path cost ; According to the inventory sufficiency, distribution path cost, and logistics load, recalculate the comprehensive recommendation score of the product. The expression is: ; Among them, represents the comprehensive score value of the product in the recommendation list, represents the comprehensive score of the product in the optimized recommendation list, represents the maximum bearable load of the logistics node, represents the logistics node at the current time point load volume, and are the weight factor of the distribution path cost and the weight factor of the logistics load respectively; According to the comprehensive score value sort the products in descending order to generate the final optimized recommendation list and display it to the user.

[0010] As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: collecting the inventory, sales volume, and logistics status data of the logistics twin model, combining with the optimized recommendation list, analyzing the product sales trend and logistics load situation, using a prediction algorithm to predict future inventory demand and logistics load, and generating an adjustment strategy. The specific steps are: Obtain inventory data, sales volume data, and logistics load data from the logistics twin model in real time; Calculate the sales trend of the product based on the collected data, with the expression: ; where represents the average sales growth rate of the product within, represents the time interval of historical data, represents the product at the current time sales data, represents the product at the past time point sales data; Based on the current inventory and sales trend, use the time series prediction algorithm to predict the inventory demand in the future for a period of time, with the expression: ; where represents the inventory level, represents the index of the product, represents the product predicted inventory level at a future time, represents the product at the current time sales data, represents the product average sales growth rate within, represents the time interval of future time; Based on the current load of the node and the order growth trend, use the time series prediction algorithm to predict the load demand in the future for a period of time, with the expression: ; where represents the predicted load of the logistics node at a future time, represents the load of the logistics node at the current time point load volume, represents the logistics node load growth rate; According to the prediction results, dynamically adjust the recommendation strategy, replenishment plan, and logistics path; When it indicates that the inventory is insufficient, and a replenishment plan is generated according to ; When it indicates that the node load is exceeded, and the node is shunted.

[0011] As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: collecting feedback data of users on the adjustment strategy and optimizing the logistics twin model through the feedback data, the specific steps are as follows: Collect user behavior data and logistics status data; Integrate the user behavior data and the logistics status data to obtain a feedback matrix ; Construct a user behavior feedback weight based on the user behavior data , measuring the overall preference of users for the recommended products ; Construct a logistics feedback weight based on the user behavior data , evaluating the adaptability of logistics nodes to provide services for products .

[0012] As a preferred solution of the intelligent e-commerce online management method described in the present invention, wherein: combining the optimization result with the adjustment strategy to dynamically adjust the recommendation strategy, replenishment plan and logistics path, the specific steps are as follows: Combine the user behavior feedback weight and the logistics feedback weight to obtain the comprehensive weight of the product in the logistics twin model, which is used to dynamically optimize the recommendation priority, and the expression is: ; wherein, represents the comprehensive feedback weight of the product , represents the set of logistics nodes related to the product ; Adjust the recommendation priority of the product according to , and the expression is: ; wherein, represents the optimized recommendation priority, represents the recommendation priority before optimization, represents the adjustment factor of the comprehensive feedback weight; Prioritize the allocation of logistics nodes with high comprehensive feedback weight to divert high-load nodes.

[0013] In a second aspect, the present invention provides an intelligent e-commerce online management system, including a collection module, a product recommendation module, a logistics twin model module, a prediction module and a dynamic adjustment module: The collection module is used to collect multi-modal data of user terminals, extract context features and generate user portraits in combination with historical data; The commodity recommendation module is used to generate a preliminary recommendation list based on the user profile, optimize the recommendation results in combination with the logistics twin model, and finally generate the final user recommendation list; The logistics twin model module is used to build a logistics twin model with candidate commodities and logistics nodes as the core, and simulate inventory, distribution paths, and logistics loads in real time; The prediction module, based on the logistics twin model data, uses prediction algorithms to analyze commodity sales trends and logistics load trends, and generates replenishment plans and logistics diversion strategies; The dynamic adjustment module collects user behavior and logistics status feedback data, calculates comprehensive weights, and dynamically optimizes recommendation strategies, replenishment plans, and logistics paths.

[0014] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the intelligent e-commerce online management method described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the intelligent e-commerce online management method described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: By collecting user terminal sensor data and operation behavior data, and combining context awareness algorithms with user historical data to dynamically generate user profiles, the present invention realizes the accurate capture of users' immediate needs. By screening preferred commodities and logistics nodes through user profiles, a logistics twin model is constructed and defined, enabling the deep integration of the recommended commodities and the status of logistics resources, and solving the problem of the separation between recommendation and logistics. Further, through the inventory status, distribution paths, and logistics load data of the logistics twin model, the recommendation list is screened and adjusted to generate an optimized recommendation list, ensuring the logistics feasibility of the recommendation results. Combining inventory, sales, and logistics status data, prediction algorithms are used to dynamically predict inventory demand and logistics load trends, and replenishment and diversion strategies are generated in advance to ensure the efficient operation of the supply chain. By collecting user feedback and logistics status data, the logistics twin model and recommendation strategies are optimized to form a feedback loop, dynamically adjusting replenishment plans, logistics paths, and recommendation priorities. The present invention realizes the collaborative optimization of commodity recommendation and logistics systems, improves recommendation accuracy, logistics efficiency, and user experience, while reducing logistics costs and resource waste, providing an efficient solution for the intelligent operation of e-commerce platforms. Description of the Drawings

[0017] 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.

[0018] Figure 1 This is a flow chart of the intelligent e-commerce online management method in Example 1.

[0019] Figure 2 This is a schematic diagram of the intelligent e-commerce online management system in Example 1. DETAILED DESCRIPTION

[0020] 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.

[0021] 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.

[0022] 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.

[0023] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, and provides an intelligent e-commerce online management method, comprising the following steps: S1. Collect sensor data and operation behavior data of user terminals, extract context data through context awareness algorithm, and combine the context data with user historical data to obtain a comprehensive feature vector; The context-aware algorithm is used to integrate the multimodal data into a context feature vector through the multimodal sensor data of the user terminal (such as obtaining geographic location information through GPS, obtaining the current ambient temperature through the temperature sensor, and capturing the user's dynamic behavior (such as walking, riding, and standing still) through the accelerometer and gyroscope) and operation behavior data (such as click records, page dwell time, and sliding behavior), combined with timestamp information (such as the current time period: morning, afternoon, and evening) and environmental context, to obtain the context feature vector , the expression is: ; Among them, represents the situational feature vector at time point , represents the geographical location feature at time point , represents the time context feature at time point , represents the environmental perception feature at time point , represents the action behavior feature at time point ; The geographical location feature is obtained through the base station positioning of the user terminal device; The time context feature is obtained through the time and calendar functions of the user device; The environmental perception feature is obtained through the built-in sensors of the user terminal or network services; The action behavior feature is obtained through the sensors of the user terminal (such as accelerometers, gyroscopes) and interaction data (such as click behavior); Collect the historical data of the user and perform frequency statistics on the number of browsing times, purchase times, and favorite times to generate the user historical feature vector ; Dynamically fuse the situational feature vector with the historical preference data of the user to generate the comprehensive feature vector , and the expression is: ; ; ; Among them, represents the comprehensive situational feature vector calculated by weighted summation from the situational feature vector , represents the weight of the th situational feature, represents at time point the th situational feature vector, represents the historical preference score calculated by weighted summation from the user historical feature vector , represents the preference score, represents the th weight of the historical feature vector, represents the th preference score of the historical feature vector represents the weight factor of the comprehensive scenario feature vector, and represents the weight factor of the historical feature vector; The comprehensive feature vector is directly mapped to the preference score of the specific commodity category through the regression model S2. According to the comprehensive feature vector, the preference commodities and logistics nodes related to the user's needs are screened out, and a logistics twin model is constructed by combining the preference commodities and logistics nodes;

[0024] Set the preference threshold as Set the preference score of the th commodity as , and screen out the commodity categories, and then perform a set operation to obtain the candidate commodity set ; ; ; Among them, represents the candidate commodity set; Each commodity category corresponds to a logistics node. Set the logistics node set as , and the logistics node related to the commodity is ; For each commodity category in the candidate commodity category set , the screening formula is: ; Among them, represents the set of logistics nodes that meet the conditions, represents the logistics node stores the inventory quantity of the commodity , represents the current logistics load of the logistics node , which is related to the distribution commodity . The larger the value, the higher the load of the node. represents the maximum bearable load of the logistics node, represents the index of the logistics node; When the inventory is insufficient, the logistics node is excluded. When the current load of the logistics node exceeds the threshold, the logistics node is excluded; The candidate commodity category set , the logistics node set and the user's delivery path set are combined to obtain the logistics twin model , and the expression is: ; According to the candidate commodity category set in the user portrait and the selected set of logistics nodes , limit the operating range of the logistics twin model, and then update the logistics twin model , to obtain the limited logistics twin model , making it only focus on the goods and related logistics nodes that the user is interested in. The expression is: ; Among them, represents the limited logistics twin model, represents the limited set of candidate product categories, represents the limited set of logistics nodes, represents the limited set of distribution paths.

[0025] S3. Generate a preliminary recommendation list based on the preferred products, input the preliminary recommendation list into the logistics twin model, and screen and adjust the recommendation list through the inventory status, distribution path, and logistics load data of the logistics twin model to generate an optimized recommendation list; Based on the set of candidate product categories , combine the sales volume factor, evaluation factor, and price factor to generate a preliminary recommendation list. The expression is: ; Among them, represents the comprehensive score of product in the preliminary recommendation list, represents the historical sales volume of product , represents the maximum historical sales volume of all products, represents the evaluation score of the user for product , represents the maximum evaluation score of all products, represents the average value of the preference scores of all products, represents the maximum preference score of all product categories, and are the weight factors of the preference score, historical sales volume, evaluation score, and price deviation respectively; The value range of is [−∞,+∞]. The higher the value, the higher the recommended priority of the product, and vice versa; Sales volume factor, indicating that products with high historical sales volume are preferably recommended; Evaluation factor, indicating that products with high user scores are preferably recommended; Price factor, indicating that products with moderate prices are preferably recommended; Input the preliminary recommendation list into the logistics twin model to screen and adjust the preliminary recommendation list; Remove the products with insufficient inventory (i.e., products with = 0); For the remaining products, calculate the distribution path cost from the logistics node to the user to obtain the total path cost. The distribution path cost formula is: For the remaining products, calculate the distribution path cost from the logistics node to the user to obtain the total path cost. The distribution path cost formula is: ; ; Among them, represents the total path cost of product delivered from all relevant logistics nodes to the user , represents the distribution distance from the logistics node to the user , represents the distribution speed of the logistics node , represents the inventory level, represents the inventory level of product in the logistics node , represents the inventory level of product in the logistics node ; According to the inventory sufficiency, distribution path cost, and logistics load, recalculate the comprehensive recommendation score of the product. The expression is: ; Among them, represents the comprehensive score value of product in the recommendation list, represents the comprehensive score of product in the optimized recommendation list, represents the maximum bearable load of the logistics node, represents the load of the logistics node at the current time point , and are the weight factors of the distribution path cost and the logistics load respectively; The value range of is [−∞, +∞], and the higher the value, the higher the recommendation priority; Sort the products in descending order according to the comprehensive score value

[0026] S4. Collect the inventory, sales volume, and logistics status data of the logistics twin model. Combine with the optimized recommendation list, analyze the commodity sales trend and logistics load situation, and use a prediction algorithm to predict future inventory requirements and logistics load Use the time series prediction algorithm to predict the inventory requirements within a future period. The expression is: ; Where, represents the inventory level, represents the index of the commodity, represents the commodity predicted inventory level at a future time, represents the commodity at the current time sales volume data, represents the commodity average sales growth rate within, represents the time interval of the future time; The value range of is [−∞, +∞]. When > 0, it means the inventory is sufficient. When = 0, it means the inventory is exhausted. When < 0, it means the inventory is insufficient and replenishment is required; Based on the current load of the node and the order growth trend, use the time series prediction algorithm to predict the load requirements within a future period. The expression is: ; Where, represents the predicted load of the logistics node at a future time, represents the load of the logistics node at the current time point , represents the load growth rate of the logistics node ; The value range of is [0, +∞]. When , it means the node load is within the acceptable range. When , it means the node load exceeds the limit and adjustment or diversion is required; According to the prediction results, dynamically adjust the recommendation strategy, replenishment plan, and logistics path; When < 0, it means the inventory is insufficient. Generate a replenishment plan according to ; When , it means the node load exceeds the limit. Diversify the node; Dynamically adjust the priority of the recommended commodities, ; Among them, represents the adjusted recommended priority of the commodity , represents the adjustment coefficient represents the maximum inventory level.

[0027] S5. Collect feedback data from users on the adjustment strategy and optimize the logistics twin model through the feedback data; Collect user behavior data (number of clicks, number of purchases, and number of abandonments) and logistics status data (actual inventory change, delivery delay time, and logistics load change); User behavior data is obtained through the platform front-end logs and user operation records; logistics status data is obtained through the real-time monitoring of the background logistics twin model; Integrate user behavior data and logistics status data to obtain a feedback matrix ; Construct a user behavior feedback weight based on user behavior data , which is used to measure the overall preference of users for the recommended commodity , and the expression is: ; Among them, represents the number of times the user clicks on the recommended commodity , represents the number of times the user purchases the recommended commodity m represents the number of times the user abandons the recommended commodity , and are the weight factors for the number of clicks, number of purchases, and number of abandonments respectively, and satisfy , ; The user behavior feedback weight , and the higher the value, the stronger the user's preference for the commodity ; When > 0, it indicates that the user has a high preference for the commodity , and the recommended weight increases; When = 0, it indicates that the user has no obvious preference for the behavior feedback of the commodity , and the original recommended weight is maintained; When < 0, it indicates that the user has a low preference for the commodity , and the recommended weight decreases; Construct a logistics feedback weight based on user behavior data , which is used to evaluate the logistics node for the commodity Adaptability of service provision, expressed as: ; where, represents the base of the natural logarithm, represents the adjustment factor for inventory changes, used to control the amount of inventory changes impact on the weight, represents the commodity inventory change amount. The higher the inventory change amount, the greater the demand for the commodity. represents the delivery delay time of the logistics node . The higher it is, the lower the weight. represents the maximum delivery delay time, represents the logistics node load change amount, represents the maximum load that the logistics node can bear; The higher the value of the more suitable the logistics node is for recommending commodities When > 0, it means that the logistics node has a high service adaptability to the commodity ; When < 0, it means that the logistics node has a low service adaptability to the commodity ;

[0028] S6. Combine the optimization results with the adjustment strategy to dynamically adjust the recommendation strategy, replenishment plan, and logistics path; Combine the user behavior feedback weight and the logistics feedback weight to obtain the comprehensive weight of the commodity in the logistics twin model, used to dynamically optimize the recommendation priority, expressed as: ; where, represents the comprehensive feedback weight of the commodity , represents the set of logistics nodes related to the commodity ; When > 0, the comprehensive recommendation priority of the commodity is high; When < 0, the comprehensive recommendation priority of the commodity is low; Adjust the recommendation priority of the commodity according to , expressed as: ; Among them, represents the optimized recommendation priority, represents the recommendation priority before optimization, represents the adjustment factor of the comprehensive feedback weight; When > 0 and the inventory is insufficient, a replenishment plan is generated : ; Among them, represents the commodity predicted inventory at a future time; Preferentially allocate the comprehensive feedback weight to logistics nodes with high weights and divert high-load nodes.

[0029] This embodiment also provides an intelligent e-commerce online management system, including: a collection module, a commodity recommendation module, a logistics twin model module, a prediction module, and a dynamic adjustment module: The collection module is used to collect multi-modal data of the user terminal, extract context features and generate a user profile in combination with historical data; The commodity recommendation module is used to generate a preliminary recommendation list according to the user profile, optimize the recommendation result in combination with the logistics twin model, and finally generate a final user recommendation list; The logistics twin model module is used to build a logistics twin model with candidate commodities and logistics nodes as the core, and real-time simulate inventory, distribution paths, and logistics loads; The prediction module, based on the logistics twin model data, uses prediction algorithms to analyze commodity sales trends and logistics load trends, and generates replenishment plans and logistics diversion strategies; The dynamic adjustment module collects user behavior and logistics status feedback data, calculates the comprehensive weight, and dynamically optimizes the recommendation strategy, replenishment plan, and logistics path.

[0030] This embodiment also provides a computer device, applicable to the case of the intelligent e-commerce online management method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent e-commerce online management method proposed in the above embodiment.

[0031] 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 implemented through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, a trackball, or a touchpad provided on the housing of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0032] This embodiment also provides a storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the method for realizing intelligent e-commerce online management proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, a magnetic disk, or an optical disc.

[0033] In summary, the present invention captures the user's immediate needs accurately by collecting the sensor data and operation behavior data of the user terminal, dynamically generating a user portrait by combining the context awareness algorithm and the user's historical data, screening the preferred products and logistics nodes through the user portrait, constructing and defining the logistics twin model, and deeply integrating the status of the recommended products and logistics resources, thus solving the problem of the disconnection between recommendation and logistics. Further, the recommended list is screened and adjusted based on the inventory status, delivery route, and logistics load data of the logistics twin model to generate an optimized recommended list, ensuring the logistics feasibility of the recommendation results. By combining the inventory, sales volume, and logistics status data, a prediction algorithm is used to dynamically predict the inventory demand and logistics load trend, and replenishment and diversion strategies are generated in advance to ensure the efficient operation of the supply chain. By collecting user feedback and logistics status data, the logistics twin model and recommendation strategy are optimized to form a feedback loop, dynamically adjusting the replenishment plan, logistics route, and recommendation priority. The present invention realizes the collaborative optimization of product recommendation and logistics system, improves the recommendation accuracy, logistics efficiency, and user experience, while reducing the logistics cost and resource waste, providing an efficient solution for the intelligent operation of e-commerce platforms.

[0034] Example 2. Referring to Table 1, this is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of the intelligent e-commerce online management method is given.

[0035] To verify the innovation and superiority of the intelligent e-commerce online management method, a set of simulation tests was designed. By comparing the operation effects of the existing technology and the method of the present invention in a specific e-commerce platform environment, the advantages of the present invention are reflected in quantitative data. The experimental environment includes a set of user terminal devices (smartphones, tablets) and an e-commerce platform server. The total number of platform products is 10,000, covering a variety of categories (electronic products, clothing, food, etc.). The total number of users is 5,000. The logistics network includes 20 warehousing and logistics nodes, and the delivery scope covers five urban areas. Geographical location, time period, ambient temperature, and user behavior data are collected through the multi-modal sensors (such as GPS, temperature sensors, accelerometers) of the user terminal, and the user's historical behavior (such as browsing records, purchase records, product ratings in the past three months, etc.) is recorded at the same time. The context awareness algorithm integrates the multi-modal sensor data through the algorithm to generate context feature vectors (such as the user is currently in the office area, the time is the lunch break, the surrounding temperature is low, and the user is stationary), and combines with the user's historical preferences to generate a user portrait. For example, a certain user portrait shows that his demand for "electronic products" and "warm clothing" is relatively high. Preferred products (such as smart watches and winter coats) are selected based on user portraits, and the inventory and load conditions of relevant logistics nodes are matched. The logistics twin model eliminates nodes with insufficient inventory or excessive logistics load. For example, a warehouse is excluded because its load exceeds 90%. The initial recommendation list is generated based on the product's preference score, sales volume, reviews, and price. It is then adjusted through the logistics twin model to remove products with insufficient inventory and calculate the delivery path cost. Finally, an optimized recommendation list is generated, such as recommending "smart watch A" and "winter coat B". Based on the recommendation list, real-time inventory, sales, and logistics load data are collected, and time series algorithms are used to predict future inventory demand and logistics load. For example, it is predicted that a certain product will be out of stock within two days and needs to be restocked in advance. Collect user clicks, purchase behaviors, and logistics feedback data (such as delivery delay time and inventory changes) to dynamically adjust recommendation strategies, replenishment plans, and logistics routes. For example, after a user purchases a recommended product, the system optimizes the load distribution of related logistics nodes.

[0036] The details are shown in Table 1 below: Table 1 Comparison of the effects of intelligent e-commerce online management methods and existing technologies

[0037] From the comparison of the tables, we can see that the intelligent e-commerce online management method is significantly better than the existing technology in many key indicators. The following is a specific analysis: Click-through rate and conversion rate of recommended products: The present invention generates dynamic user portraits, combines contextual awareness and historical preferences, accurately matches user needs, and significantly improves the click-through rate (from 15.2% to 28.7%, an increase of 88.8%) and conversion rate (from 6.3% to 14.1%, an increase of 123.8%) of recommended products. Traditional recommendation systems lack real-time contextual awareness capabilities, and their recommendation results easily deviate from users' immediate needs.

[0038] Average delivery time: The present invention screens and adjusts the recommendation results through the logistics twin model, giving priority to recommending products with high logistics feasibility, and dynamically optimizing the logistics route, so that the average delivery time is reduced from 36 hours to 28 hours, a reduction of 22.2%. The traditional logistics system fails to link with the recommendation system, resulting in low delivery efficiency.

[0039] Inventory out-of-stock rate: By adopting a predictive algorithm, replenishing stocks in advance and optimizing inventory allocation, the present invention reduces the inventory out-of-stock rate from 10% to 3.2%, a reduction of 68%. Traditional methods fail to effectively predict inventory demand and are prone to out-of-stock problems.

[0040] User recommendation satisfaction score: The present invention combines recommendation optimization with dynamic adjustment of logistics, significantly improving the user experience and increasing the satisfaction score from 3.8 to 4.6 (a 21.1% increase). The traditional recommendation system has a low user satisfaction due to untimely logistics response.

[0041] Average load rate of logistics nodes: The present invention reduces the load rate of logistics nodes from 82% to 68% (a 17% reduction) by diverting high-load nodes and dynamically adjusting logistics routes. The traditional logistics system has uneven load distribution, which easily leads to overload of some nodes.

[0042] Average operating cost: The present invention reduces inventory waste and distribution costs through collaborative optimization of recommendation and logistics, reducing the average cost per order from 12.5 yuan to 9.2 yuan (a 26.4% reduction). The traditional method has low resource utilization due to the lack of a collaborative mechanism.

[0043] In summary, through multi-link innovation, the present invention not only significantly improves the accuracy of the recommendation system and the efficiency of the logistics system, but also effectively reduces the operating cost of the platform, fully demonstrating its creativity and practicality in intelligent e-commerce online management.

[0044] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not restrictive. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. An intelligent e-commerce online management method, characterized by: include: Collect sensor data and operation behavior data from user terminals, extract context data through context awareness algorithms, and combine the context data with user historical data to obtain a comprehensive feature vector; According to the comprehensive feature vector, the preferred products and logistics nodes related to user needs are screened out, and the logistics twin model is constructed by combining the preferred products and logistics nodes; Generate a preliminary recommendation list based on preferred products, input the preliminary recommendation list into the logistics twin model, filter and adjust the recommendation list through the inventory status, delivery route and logistics load data of the logistics twin model, and generate an optimized recommendation list; Collect inventory, sales, and logistics status data from the logistics twin model, combine it with the optimized recommendation list, analyze product sales trends and logistics load, use prediction algorithms to predict future inventory demand and logistics load, and generate adjustment strategies; Collect user feedback data on adjustment strategies and optimize the logistics twin model through feedback data; Combine optimization results with adjustment strategies to dynamically adjust recommendation strategies, replenishment plans, and logistics routes.

2. The intelligent e-commerce online management method according to claim 1, characterized in that: The sensor data and operation behavior data of the user terminal are collected, the context data is extracted through the context perception algorithm, and the context data is combined with the user's historical data to obtain a comprehensive feature vector. The specific steps are as follows: Through the multimodal sensor data and operation behavior data of the user terminal, combined with timestamp information and environmental context, the context-aware algorithm is used to integrate the multimodal data into a context feature vector to obtain the context feature vector ; Collect user historical data and perform frequency statistics on browsing times, purchase times, and collection times to generate user historical feature vectors ; The situation feature vector Dynamically integrate with the user's historical preference data to generate a comprehensive feature vector , the expression is: ； ； ； in, Represents the context feature vector The comprehensive situational feature vector obtained by weighted summation in Indicates The weight of the situational features, Indicates at a point in time No. A context feature vector, Represents the historical feature vector from the user The historical preference score calculated by weighted summation in Indicates the preference score, Indicates The weight of the historical feature vectors, Indicates The preference scores of historical feature vectors, represents the weight factor of the comprehensive situational feature vector, Represents the weight factor of the historical feature vector; The comprehensive feature vector is transformed into Directly mapped to preference scores for specific product categories.

3. The intelligent e-commerce online management method according to claim 2, characterized in that: According to the comprehensive feature vector, the preferred products and logistics nodes related to user needs are screened out, and the logistics twin model is constructed by combining the preferred products and logistics nodes. The specific steps are as follows: Set the preference threshold to , set the The preference score of a product is , filter out The product categories are then aggregated to obtain a candidate product set ; Each commodity category corresponds to a logistics node, and the logistics node set is set as , and commodities The relevant logistics nodes are ; For the candidate product category set Filter each product category in; When the inventory is insufficient, the logistics node is eliminated; when the current load of the logistics node exceeds the threshold, the logistics node is eliminated; Set the candidate product categories , Logistics Node Collection and the user's delivery path collection Collect and obtain the logistics twin model .

4. The intelligent e-commerce online management method according to claim 3, characterized in that: The method generates a preliminary recommendation list based on the preferred commodities, inputs the preliminary recommendation list into the logistics twin model, screens and adjusts the recommendation list through the inventory status, delivery path and logistics load data of the logistics twin model, and generates an optimized recommendation list. The specific steps are as follows: In the candidate product category set On this basis, a preliminary recommendation list is generated by combining sales volume factor, evaluation factor and price factor. The expression is: ； in, Display products The overall score in the preliminary recommendation list, Display products Historical sales volume, Indicates the maximum historical sales volume of all products. Indicates the user's The evaluation score of Indicates the maximum evaluation score of all products. represents the average preference score of all products. represents the maximum preference score of all product categories, and They are the weight factors of preference score, historical sales volume, evaluation score and price deviation; A preliminary list of recommendations Enter the Logistics Twin Model Screen and adjust the preliminary recommendation list; Eliminate out-of-stock items; For the remaining goods, calculate their logistics nodes The delivery path cost to the user, the total path cost ; According to inventory adequacy, delivery path cost and logistics load, the comprehensive recommendation score of the product is recalculated as follows: ； in, Display products The comprehensive rating value in the recommendation list, Display products The comprehensive rating in the optimization recommendation list, Indicates the maximum load that a logistics node can bear. Represents a logistics node At the current time The load capacity, and They are the weight factors of the distribution path cost and the logistics load respectively; According to the comprehensive rating Arrange the products in descending order, generate the final optimized recommendation list, and display it to the user.

5. The intelligent e-commerce online management method according to claim 4, characterized in that: The inventory, sales volume and logistics status data of the logistics twin model are collected, combined with the optimized recommendation list, the commodity sales trend and logistics load are analyzed, the prediction algorithm is used to predict future inventory demand and logistics load, and an adjustment strategy is generated. The specific steps are as follows: Obtain inventory data, sales data, and logistics load data in real time from the logistics twin model; The sales trend of the goods is calculated based on the collected data. The expression is: ； in, Display products The average sales growth rate within Indicates the time interval of historical data, Display products At current time Sales data, Display products At a point in the past Sales data; Based on the current inventory and sales trends, the time series forecasting algorithm is used to predict the inventory demand in the future. The expression is: ； in, Indicates the inventory quantity. Indicates the index of the product. Display products The estimated inventory level at a future time, Display products At current time Sales data, Display products The average sales growth rate within A time interval representing a future time; Based on the current node load and order growth trend, the time series prediction algorithm is used to predict the load demand in the future. The expression is: ； in, Represents a logistics node The expected load at a future time, Represents a logistics node At the current time The load capacity, Represents a logistics node The load growth rate; Dynamically adjust recommendation strategies, replenishment plans, and logistics routes based on forecast results; when When the inventory is insufficient, Generate replenishment plans; when , it indicates that the node load exceeds the limit and the node is diverted.

6. The intelligent e-commerce online management method according to claim 5, characterized in that: The specific steps of collecting user feedback data on the adjustment strategy and optimizing the logistics twin model through the feedback data are as follows: Collect user behavior data and logistics status data; Integrate user behavior data and logistics status data to obtain a feedback matrix ; Constructing user behavior feedback weights based on user behavior data , which measures the user's preference for recommended products Overall preference; Constructing logistics feedback weights based on user behavior data , evaluate logistics nodes For goods Provide adaptability of services.

7. The intelligent e-commerce online management method according to claim 6, characterized in that: The optimization results are combined with the adjustment strategy to dynamically adjust the recommendation strategy, replenishment plan and logistics path. The specific steps are as follows: Weighting user behavior feedback and logistics feedback weight Fusion, get goods The comprehensive weight in the logistics twin model is used to dynamically optimize the recommendation priority, and the expression is: ； in, Display products The comprehensive feedback weight of Display and products A collection of related logistics nodes; according to Adjust the recommendation priority of the product, the expression is: ； in, Indicates the optimized recommendation priority. Indicates the recommendation priority before optimization. represents the adjustment factor of the comprehensive feedback weight; Prioritize the allocation of comprehensive feedback weight High logistics nodes, divert high load nodes.

8. An intelligent online e-commerce management system, based on the intelligent online e-commerce management method according to any one of claims 1 to 7, characterized in that: Including collection module, product recommendation module, logistics twin model module, prediction module and dynamic adjustment module: The acquisition module is used to collect multimodal data of user terminals, extract situational features and generate user portraits in combination with historical data; The product recommendation module is used to generate a preliminary recommendation list based on the user portrait, optimize the recommendation results in combination with the logistics twin model, and finally generate a final user recommendation list; The logistics twin model module is used to build a logistics twin model with candidate commodities and logistics nodes as the core, and simulate inventory, distribution routes and logistics loads in real time; The prediction module, based on the logistics twin model data, uses the prediction algorithm to analyze the commodity sales trend and logistics load trend, and generates a replenishment plan and logistics diversion strategy; The dynamic adjustment module collects user behavior and logistics status feedback data, calculates comprehensive weights, and dynamically optimizes recommendation strategies, replenishment plans, and logistics routes.

9. 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 intelligent e-commerce online management method described in any one of claims 1 to 7 are implemented.

10. 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 intelligent e-commerce online management method described in any one of claims 1 to 7 are implemented.

Images