E-commerce marketing optimization system based on AI

Abstract

The invention relates to the technical field of marketing data analysis, in particular to an AI-based e-commerce marketing optimization system, and the system comprises a user behavior analysis module which collects the behavior data of a user on an e-commerce platform, divides the behavior data of the user on the e-commerce platform into a plurality of behavior categories, and forms a user behavior statistical result. According to the method, through collection, classification, statistics and analysis of the user behavior data, the characteristics of behaviors such as browsing, selection and purchasing of the user can be accurately obtained, so that the user behavior categories are refined, and the classification data are generated. The refinement is helpful for deeply mining user behavior modes and improving the pertinence of marketing strategies. In combination with cross-platform behavior path analysis and distribution characteristics of active time periods, behavior chains of users on different platforms can be effectively tracked, a more comprehensive user behavior view is provided, and optimization of cross-platform resource allocation and path planning is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of marketing data analysis, and in particular to an AI-based e-commerce marketing optimization system. Background Art

[0002] Marketing data analysis improves the efficiency and effectiveness of marketing by analyzing large amounts of marketing data. The field uses various data analysis tools and technologies, such as data mining, machine learning, and big data platforms, to identify consumer preferences, evaluate the performance of marketing activities, and adjust marketing strategies based on this information. The data analyzed can include sales data, customer feedback, online behavior records, and social media interactions.

[0003] Among them, the e-commerce marketing optimization system helps merchants optimize their marketing strategies and improve sales efficiency by integrating and analyzing sales data, customer behavior data and market trends on e-commerce platforms. The main uses of the system include automatic product recommendations, personalized marketing campaign design, price optimization, inventory management, and customer relationship maintenance, enabling merchants to more accurately target customer groups and effectively improve user purchase rates and customer satisfaction.

[0004] Existing technologies in user behavior analysis mostly stay at the basic data statistics level, lacking in-depth exploration of behavior classification and its dynamic changes, resulting in ambiguity and incompleteness in the acquisition of user behavior characteristics. The analysis methods for cross-platform user behavior are single, making it difficult to effectively identify the behavioral linkage and active periods of users across multiple platforms, and unable to provide strong support for resource optimization and allocation. In terms of order processing, only simple summarization of payment status is focused on, and in-depth analysis of the reasons for order cancellation and user group distribution is not conducted, which can easily lead to abnormal order problems not being discovered and resolved in a timely manner, wasting corporate resources. For user stratification and behavior prediction, existing technologies often use static models, which cannot capture the dynamic characteristics of behavior paths and cannot make accurate predictions of user transaction behaviors. In addition, existing technologies lack detailed analysis in key node optimization, making it difficult to adjust marketing path design according to user behavior characteristics, thereby reducing conversion rates and user satisfaction. Summary of the invention

[0005] In order to solve the technical problems existing in the prior art, an embodiment of the present invention provides an e-commerce marketing optimization system based on AI, and the system includes: The user behavior analysis module collects the user's behavior data on the e-commerce platform, divides the user's behavior data on the e-commerce platform into multiple behavior categories, and forms user behavior statistics; The cross-domain behavior analysis module analyzes the user's cross-platform behavior path based on the user behavior statistics, selects the priority access path, analyzes the active time period and path of the user's behavior on the platform, and generates the user's cross-platform path distribution data; The abnormal order identification module obtains the operation records of the order payment status in the user behavior statistics, marks the users who cancel the orders for multiple times in a row, and obtains the abnormal order data analysis results; The user stratification and dynamic feature extraction module identifies the change trend of order cancellation behavior in each access category target time period based on the user's cross-platform path distribution data and abnormal order data analysis results, and generates a stratified user feature extraction result; The user transaction pattern analysis module analyzes the user's behavior chain on the access path based on the hierarchical user feature extraction result, determines the stability of the behavior chain within the path, and builds a user transaction behavior prediction model; The e-commerce marketing data optimization module classifies the user's future operation tendencies at key behavior nodes through the user transaction behavior prediction model, cross-analyzes the associated features of abnormal behaviors and unfinished transaction paths, adjusts the behavior chain data structure, and generates marketing behavior chain optimization data.

[0006] The present invention has the following improvements: the user behavior statistics include statistics of the user's page browsing time, analysis results of product click-through rates, and distribution of types and quantities of products added to the shopping cart; the user cross-platform path distribution data include cumulative results of the number of user visits to the platform, path classification statistics of behavior jump times, and ranking results of path visit times; the abnormal order data analysis results include distribution of order cancellation rates for each user, a list of marked users who continuously cancel orders, and order records associated with canceled orders; the hierarchical user feature extraction results include user distribution statistics classified by single-path access and multi-path access, data analysis of priority access path concentration, and distribution trends of users in active time periods; the user transaction behavior prediction model includes analysis results of node stay time in a single-path access user behavior chain, and stability data of behavior sequence; the marketing behavior chain optimization data include user behavior feature comparison analysis results, evaluation data of the degree of influence of path nodes on the behavior chain, and user behavior chain optimization after critical path adjustment.

[0007] The present invention is improved in that the user behavior analysis module comprises: The user behavior data collection submodule collects the user's page browsing time, product click-through rate, type and quantity of products added to the shopping cart, order payment status and product access frequency based on the user's behavior data on the e-commerce platform, and generates the original user behavior data; The user behavior classification and sorting submodule divides the user behavior into multiple categories such as browsing, selecting, and purchasing based on the original user behavior data, and sorts them by platform and user behavior type to generate user behavior classification data; The user behavior statistics submodule counts the key indicators of each category of user behavior based on the user behavior classification data, including the total browsing behavior duration, selection behavior frequency, and order payment ratio of purchase behavior, and generates user behavior statistics results.

[0008] The present invention is improved in that the cross-domain behavior analysis module includes: The cross-platform behavior extraction submodule extracts the number of visits and behavior jumps of users between multiple platforms based on the user behavior statistics, analyzes the jump frequency of the behavior path, and generates user cross-platform access path data; The path frequency statistics submodule summarizes the behavior jump frequency according to the path based on the user's cross-platform access path data, obtains the visit frequency ratio of each path, sorts the path visit frequency, and generates path access frequency data; The user active distribution analysis submodule analyzes the user's active time periods and path characteristics across multiple platforms based on the path access frequency data and combined with the classified statistics of the time periods in which the user's behavior occurs, and generates user cross-platform path distribution data.

[0009] The present invention is improved in that the abnormal order identification module comprises: The order record processing submodule extracts the payment success record and payment cancellation record based on the operation record of the order payment status in the user behavior statistics, counts the number of order submissions and cancellations for each user, and generates user order record statistics; The order cancellation rule analysis submodule obtains the order cancellation rate of each user based on the statistical data of the user order records, counts the number and proportion of users within the cancellation rate range, analyzes the distribution of the order cancellation rate in the user group, and generates an order cancellation rate analysis result; The abnormal order marking submodule marks users who cancel orders multiple times in a row based on the order cancellation rate analysis results, extracts order records associated with the cancellation behavior, and generates abnormal order data analysis results.

[0010] The present invention is improved in that the user stratification and dynamic feature extraction module includes: The user behavior path stratification submodule stratifies the user behavior paths according to the concentration of the priority access paths and the distribution of users in the active time period based on the user cross-platform path distribution data, and generates user behavior path stratification data; The path cancellation rate analysis submodule generates path cancellation rate distribution data based on the analysis results of the user behavior path hierarchical data and abnormal order data and the path access category, and calculates the order cancellation rate distribution of single-path access and multi-path access users; The behavior trend identification submodule analyzes the dynamic change characteristics of the order cancellation behavior based on the path cancellation rate distribution data and the change trend of the order cancellation behavior in the target time period, and generates a hierarchical user feature extraction result.

[0011] The present invention is improved in that, for analyzing the dynamic change characteristics of order cancellation behavior, according to the formula: ; Calculation time period Weighted average cancellation rate ;in, Is a user In time period The order cancellation rate Is a user The path complexity weight of It is the total number of all users on the e-commerce platform.

[0012] The present invention is improved in that the user transaction pattern analysis module comprises: The single-path behavior chain analysis submodule analyzes the behavior path of the single-path access user based on the hierarchical user feature extraction result, identifies the sequence and stay characteristics of the behavior nodes, determines the stability of the behavior chain, and generates single-path behavior chain analysis data; The multi-path behavior linkage analysis submodule analyzes the cross-path behavior distribution of multi-path access users based on the single-path behavior chain analysis data, determines the behavior linkage characteristics between associated paths, and generates multi-path behavior linkage analysis data; The user behavior prediction model building submodule calculates the conversion probability of the behavior node based on the multi-path behavior linkage analysis data, and builds a user transaction behavior prediction model according to the behavior chain characteristics from browsing to purchasing and the user's behavior on the access path.

[0013] The present invention is improved in that the conversion probability of the behavior node is calculated by using the formula: ; Get the slave behavior node Convert to Node The conversion probability ; in, It is a slave node Convert to Node The total number of times, Is a node The average length of stay, Is a node Total visits, Is a node average stay time.

[0014] The present invention is improved in that the e-commerce marketing data optimization module includes: The key behavior node classification submodule classifies the user's operation tendency at the key behavior nodes based on the user transaction behavior prediction model, extracts the user behavior data of completed transaction operations and the user behavior characteristics of uncompleted transaction operations in the key path, and generates key node user behavior data; The path node impact analysis submodule analyzes the impact of the path node on the user behavior chain based on the key node user behavior data, cross-analyzes the correlation characteristics of abnormal behavior and unfinished transaction paths, and generates path node impact analysis data; The marketing behavior chain adjustment submodule adjusts the behavior chain data structure based on the path node impact analysis data, optimizes the node layout of the key path, and generates marketing behavior chain optimization data.

[0015] The beneficial effects brought by the technical solution provided by the embodiment of the present invention include at least: by collecting, classifying, counting and analyzing user behavior data, the characteristics of user browsing, selecting, purchasing and other behaviors can be accurately obtained, thereby refining the user behavior category and generating classified data. This refinement helps to deeply explore user behavior patterns and improve the pertinence of marketing strategies. Combining the cross-platform behavior path analysis and the distribution characteristics of active time periods, it is possible to effectively track the behavior chain of users on different platforms, provide a more comprehensive user behavior view, and help optimize cross-platform resource allocation and path planning. The in-depth analysis of order payment status and cancellation records, by extracting the distribution law of cancellation rate and marking abnormal orders, not only improves data quality, but also can detect potential problem users early and reduce resource waste. Through the hierarchical analysis of path access frequency and cancellation rate, the accuracy of user behavior feature extraction is significantly optimized, supporting the implementation of more accurate marketing plans for different user groups. In transaction behavior analysis, combined with the linkage characteristics of single-path and multi-path behavior chains, the key nodes in the user behavior chain and their influence can be more deeply identified, thereby building an accurate behavior prediction model. Ultimately, by optimizing the layout of key behavioral nodes and paths, we were able to effectively guide user transaction behavior, significantly improving user conversion rates and customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a system module diagram of the present invention; Figure 2 It is a system framework diagram of the present invention; Figure 3 is a schematic diagram of a user behavior analysis module of the present invention; Figure 4 is a schematic diagram of a cross-domain behavior analysis module of the present invention; Figure 5 Schematic diagram of an abnormal order identification module of the present invention; Figure 6 A schematic diagram of a user stratification and dynamic feature extraction module of the present invention; Figure 7 A schematic diagram of a user transaction pattern analysis module of the present invention; Figure 8 Schematic diagram of the e-commerce marketing data optimization module of the present invention. DETAILED DESCRIPTION

[0018] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0019] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "example" in the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either of the two.

[0020] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference between them is not emphasized, the meanings they intend to express are the same.

[0021] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0022] In order to make the technical problems, technical solutions and advantages to be solved by the present invention more clear, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0023] See also Figure 1 , an embodiment of the present invention provides an AI-based e-commerce marketing optimization system, the system comprising: The user behavior analysis module collects the user's behavior data on the e-commerce platform, divides the user's behavior data on the e-commerce platform into multiple behavior categories, and forms user behavior statistics; The cross-domain behavior analysis module analyzes the user's cross-platform behavior path based on the user behavior statistics. By selecting the priority access path, it analyzes the active time period and path of the user's behavior on the platform and generates the user's cross-platform path distribution data; The abnormal order identification module obtains the operation records of order payment status in the user behavior statistics, marks users who cancel orders multiple times in a row, and obtains abnormal order data analysis results; The user stratification and dynamic feature extraction module identifies the changing trend of order cancellation behavior in each access category target time period based on the user's cross-platform path distribution data and abnormal order data analysis results, and generates stratified user feature extraction results; The user transaction pattern analysis module analyzes the user's behavior chain on the access path based on the hierarchical user feature extraction results, determines the stability of the behavior chain within the path, and builds a user transaction behavior prediction model; The e-commerce marketing data optimization module classifies users' future operation tendencies at key behavior nodes through the user transaction behavior prediction model, cross-analyzes the correlation characteristics of abnormal behaviors and unfinished transaction paths, adjusts the behavior chain data structure, and generates marketing behavior chain optimization data.

[0024] The user behavior statistics include the user's page browsing time statistics, product click-through rate analysis results, and the distribution of the types and quantities of products added to the shopping cart. The user's cross-platform path distribution data includes the cumulative results of the user's platform visits, the path classification statistics of the number of behavior jumps, and the ranking results of the path visit ratio. The abnormal order data analysis results include the distribution of each user's order cancellation rate, the marked list of users who continuously canceled orders, and the order records associated with the canceled orders. The stratified user feature extraction results include the user distribution statistics classified by single-path access and multi-path access, the data analysis of the concentration of priority access paths, and the distribution trend of users in active time periods. The user transaction behavior prediction model includes the analysis results of the node stay time in the single-path access user behavior chain and the stability data of the behavior sequence. The marketing behavior chain optimization data includes the user behavior feature comparison analysis results, the evaluation data of the impact of path nodes on the behavior chain, and the user behavior chain optimization after the key path adjustment.

[0025] See also Figure 2 and Figure 3 , the user behavior analysis module includes: The user behavior data collection submodule collects the user's page browsing time, product click-through rate, type and quantity of products added to the shopping cart, order payment status and product access frequency based on the user's behavior data on the e-commerce platform, and generates the original user behavior data; The user's behavior data on the e-commerce platform, including page browsing time, product click-through rate, types and quantity of products added to the shopping cart, order payment status and product access frequency, is first obtained by recording the behavior logs of the e-commerce platform and continuously monitored by the real-time acquisition module of the log system. The page browsing time is calculated based on the time difference between the timestamp when the user enters the page and the timestamp when the user leaves the page. The product click-through rate is obtained by calculating the ratio of the total number of views of the product to the number of user clicks. The types and quantities of products added to the shopping cart are obtained by parsing the operation records of the user's shopping cart to obtain the names and corresponding quantities of different products. The order payment status is determined by the transaction status mark of the payment operation, including payment success, payment failure, payment cancellation and other states. The product access frequency is measured by counting the number of visits to the product page and normalizing it according to a specific time period to measure the user's interest in a specific product. All the acquired data is structured and stored through the data storage module, with a unified processing format and removal of abnormal and duplicate records, and finally integrated to obtain the original user behavior data.

[0026] The user behavior classification submodule divides user behaviors into multiple categories such as browsing, selecting, and purchasing based on the original user behavior data, and organizes them by platform and user behavior type to generate user behavior classification data; Based on the original data of user behavior, user behavior is divided into multiple categories such as browsing, selecting, and purchasing. First, the browsing category is determined based on whether the user only stays on the page and does not generate other operations. The user's stay time on the page is recorded, and no operations such as clicking on product details, adding to shopping carts, or placing orders are performed. This behavior is classified as a browsing category. For the selection category, first, it is judged by analyzing whether the user clicks on product details or performs the add to shopping cart operation. If the user stays on the product details page and has a clear selection operation, such as adding to the shopping cart or viewing related recommendations, this behavior is classified as a selection category. For the purchase category, it is confirmed based on the user's order payment status on the e-commerce platform. If the user successfully completes the payment operation, the behavior is classified as a purchase category. All category divisions are classified and sorted according to the e-commerce platform and user behavior type, and duplicate and abnormal data in the behavior are eliminated and corrected through data cleaning to ensure the accuracy and consistency of the classified data, and finally generate user behavior classification data.

[0027] The user behavior statistics submodule is based on user behavior classification data, and counts the key indicators of each category of user behavior, including the total browsing behavior duration, selection behavior frequency, and order payment ratio of purchase behavior, to generate user behavior statistics results; Based on the user behavior classification data, key indicators are extracted for each category of behavior data. First, the total browsing behavior duration is calculated. The total browsing behavior duration is obtained by cumulatively adding up the time users stay on the e-commerce page, ensuring that the length of time users stay on the same page for multiple visits is included in the statistics. For the calculation of the selection behavior frequency, the number of times users click on product details or add products to the shopping cart is counted and the total frequency of selection behavior is calculated, where each click on product details or addition to the shopping cart is regarded as an occurrence of selection behavior. Finally, the order payment ratio of purchase behavior is calculated based on the ratio of the number of orders that users successfully pay to the total number of orders. The order payment ratio is obtained by classifying and counting the payment status of each user's order, and calculating the ratio of successfully paid orders to the total number of orders to measure the user's purchasing behavior tendency. All statistical results are classified and stored by category and the data is formatted during the final collation to ensure that the generated statistical results can be directly applied to subsequent user behavior analysis, and finally generate user behavior statistical results.

[0028] See also Figure 2 and Figure 4 , the cross-domain behavior analysis module includes: The cross-platform behavior extraction submodule extracts the number of user visits and behavior jumps between multiple platforms based on user behavior statistics, analyzes the jump frequency of the behavior path, and generates user cross-platform access path data; Based on the statistical results of user behavior, the number of visits and behavior jumps between multiple platforms are extracted. The number of visits is obtained by counting the user's login records on each platform. Each login event is assigned to the corresponding platform according to the timestamp. The number of jumps is determined by analyzing the user's behavior sequence, where each behavior sequence represents all operations of the user during a login. The number of platform switching is counted by the change of the platform identifier in the behavior sequence. The generation of the behavior sequence is arranged in chronological order, and the user's behavior path is organized into a timeline structure. The structure is used to extract the number of user behavior jumps between platforms. Finally, all the visit and jump data are stored in a unified table and classified by user to generate user cross-platform access path data.

[0029] The path frequency statistics submodule summarizes the behavior jump frequency by path based on the user's cross-platform access path data, sorts the path access times according to the proportion of each path's access times, and generates path access frequency data; Based on the user's cross-platform access path data, the behavior jump frequency is summarized according to the path. The jump frequency is calculated by dividing the cumulative number of jumps of the path by the total number of visits to the path. The path visit number data is generated by statistically analyzing the user's cross-platform access path table, where each path represents a jump behavior between a pair of platforms. The number of visits is summarized by the path identifier and the total number of visits of all paths is counted. The access proportion of the path is normalized, and finally the path access proportion is arranged in descending order to generate a sorting result, which is stored in a structured data table to generate path access frequency data.

[0030] The user active distribution analysis submodule analyzes the user's active time periods and path characteristics across multiple platforms based on the path access frequency data and the classified statistics of the time periods in which the user's behavior occurs, and generates the user's cross-platform path distribution data; Based on the path access frequency data and combined with the classified statistics of the time periods when user behaviors occur, the time periods when user behaviors occur are divided according to the timestamp information recorded in the logs. A day is divided into different time periods such as early morning, morning, afternoon, and evening. The time when the behavior occurs is mapped to the corresponding time period and the number of behaviors in each time period is counted. The access data of each path and the time period distribution data are cross-analyzed to evaluate the changing trends of users' access behaviors on each path over time, extract the distribution characteristics of active users in cross-platform paths, and finally generate a comprehensive table of time periods and path behavior data to generate users' cross-platform path distribution data.

[0031] See also Figure 2 and Figure 5 ,The abnormal order identification module includes: The order record processing submodule extracts the payment success record and payment cancellation record based on the operation record of the order payment status in the user behavior statistics, counts the number of order submissions and cancellations for each user, and generates user order record statistics; Based on the operation records of order payment status in the user behavior statistics, successful payment records and cancellation payment records are extracted. Successful payment records are obtained by filtering data with transaction status marked as "paid", and cancellation payment records are obtained by filtering data with transaction status marked as "canceled". The order submission operation records of each user are grouped, and the total number of order submissions and cancellations of each user are counted. The number of order submissions is determined by counting all order items in the operation records, and the number of cancellations is obtained by counting order items in the "canceled" status. Finally, the order statistics data of each user are integrated into a structured data table, and the data in the table is sorted according to the user identifier to generate user order record statistics.

[0032] The order cancellation pattern analysis submodule obtains the order cancellation rate of each user based on the statistical data of user order records, counts the number and proportion of users within the cancellation rate range, analyzes the distribution pattern of the order cancellation rate in the user group, and generates the order cancellation rate analysis results; Based on the statistical data of user order records, the order cancellation rate of each user is obtained. The cancellation rate is calculated as the ratio of the number of user order cancellations to the total number of submissions. After the cancellation rates of all users are calculated through batch processing, the results are divided into several intervals, such as 0-10%, 10-30%, etc. The number of users in each interval is counted and their proportion is calculated. The proportion is the ratio of the number of users in this interval to the total number of users. Finally, a user distribution table is generated in each cancellation rate interval. The characteristics of different cancellation rates in the user group are analyzed in combination with the distribution law to generate the order cancellation rate analysis results.

[0033] The abnormal order marking submodule marks users who have canceled orders multiple times in a row based on the order cancellation rate analysis results, extracts order records associated with the cancellation behavior, and generates abnormal order data analysis results; Based on the order cancellation rate analysis results, users who cancel orders multiple times in a row are marked. The marking of consecutive order cancellations is achieved by analyzing the transaction status of adjacent operations in the user's order records. Users who are marked as "canceled" twice or more in a row are marked as high-cancellation risk users. Order records associated with cancellation behavior are extracted from high-cancellation risk users, including order creation time, cancellation reason, product category and other information. The user's behavior patterns and preferences are further analyzed through the extracted associated order information, and the results are stored in the abnormal order database to generate abnormal order data analysis results.

[0034] See also Figure 2 and Figure 6 , the user stratification and dynamic feature extraction module includes: The user behavior path stratification submodule stratifies the user behavior path according to the concentration of the priority access path and the distribution of users in the active time period based on the user's cross-platform path distribution data, and generates user behavior path stratification data; Based on the user's cross-platform path distribution data, the user behavior path is stratified according to the concentration of priority access paths and the distribution of users in active time periods. The concentration of priority access paths is calculated by counting the ratio of the number of users' visits to specific paths to their total number of visits. The distribution of active time periods is calculated by dividing the time when user behaviors occur according to fixed time periods, and counting the frequency of user behaviors in each time period. The above statistical results are combined with the user's path feature data to stratify the users' behaviors. After stratification, the users are classified and sorted into single-path and multi-path categories. Finally, the stratification results are stored to generate user behavior path stratification data.

[0035] The path cancellation rate analysis submodule is based on the analysis results of user behavior path hierarchical data and abnormal order data, combined with path access categories, to calculate the order cancellation rate distribution of single-path access and multi-path access users, and generate path cancellation rate distribution data; Based on the analysis results of user behavior path hierarchical data and abnormal order data, combined with path access categories, the order cancellation rate distribution of single-path access and multi-path access users is statistically analyzed. Single-path access users are identified by judging whether they have only one continuous behavior path on the e-commerce platform, and multi-path access users are confirmed by counting the records involving multiple platforms or path jumps in the behavior sequence. The order cancellation rate distribution is calculated by calculating the average cancellation rate of users under each path category and classifying them. The classification result uses the path category as the primary key, and the user behavior cancellation rate is stored as a structured table as a statistical value to ensure that the order cancellation behavior of single path and multi-path can be distinguished, and finally the path cancellation rate distribution data is generated.

[0036] The behavior trend identification submodule is based on the path cancellation rate distribution data and the changing trend of order cancellation behavior in the target time period, analyzes the dynamic change characteristics of order cancellation behavior, and generates hierarchical user feature extraction results; To analyze the dynamic characteristics of order cancellation behavior, according to the formula: ; Calculation time period Weighted average cancellation rate ; in, Is a user In time period The order cancellation rate, measured by the number of order cancellations and total order quantity The ratio of is obtained: , and Directly extracted from the transaction records of the e-commerce platform, Is a user The path complexity weight reflects the complexity of its path behavior, through the number of paths and jump count Weighted and normalized calculation: , Is a user The number of access paths is obtained by counting the number of unique path identifiers in the log records. Is a user The number of path jumps is obtained by analyzing the path change events in the user behavior sequence. and is the maximum value of the number of paths and jumps among all users, used for normalization. It is the total number of all users on the e-commerce platform.

[0037] For example, a target time period , there are 3 users in total. The number of order cancellations and total order numbers of the users are (2, 10), (3, 10), and (1, 10) respectively. The number of paths is (2, 5, 1), the number of jumps is (3, 7, 1), the maximum number of paths is 5, and the maximum number of jumps is 7: Calculate the cancellation rate for each user: ; Calculate the path complexity weight of each user: ; Calculate the weighted average cancellation rate: ; Assuming that the weighted average cancellation rate of the previous cycle is 0.25, the result of the current example is 0.22, indicating that the order cancellation behavior in the target time period shows a downward trend. Combined with the path complexity weight, it can be seen that multi-path access users have a greater impact on cancellation behavior. The path design or strategy adjustment can be further optimized for multi-path access users, and finally the hierarchical user feature extraction results are generated.

[0038] See also Figure 2 and Figure 7 , the user transaction pattern analysis module includes: The single-path behavior chain analysis submodule analyzes the behavior path of single-path access users based on the hierarchical user feature extraction results, identifies the sequence and stay characteristics of the behavior nodes, determines the stability of the behavior chain, and generates single-path behavior chain analysis data; Based on the results of hierarchical user feature extraction, the behavior path of single-path access users is analyzed to identify the order and stay characteristics of behavior nodes. Behavior nodes are extracted through the user's operation log on the page, including key nodes such as visiting the page, clicking on product details, and adding to the shopping cart. The node order is obtained by sorting the user's operation time on the page. The stay characteristics are obtained by analyzing the length of time the user stays at each node. The length of stay time is extracted from the timestamp data of entering and leaving the node. The stability of the behavior chain is obtained by analyzing the consistency of access patterns and the degree of behavior repetition between nodes. Finally, the data table of the single-path behavior chain is integrated, combined with the characteristic analysis of stay time and access sequence, to generate single-path behavior chain analysis data.

[0039] The multi-path behavior linkage analysis submodule analyzes the cross-path behavior distribution of multi-path access users based on the single-path behavior chain analysis data, determines the behavior linkage characteristics between associated paths, and generates multi-path behavior linkage analysis data; Based on the single-path behavior chain analysis data, the cross-path behavior distribution of multi-path access users is analyzed. The multi-path behavior sequence is obtained through the user's access and jump records on multiple e-commerce platforms. The distribution characteristics of each path are extracted through the number of user operation behaviors in the path and the path residence time. The correlation characteristics between paths are obtained by analyzing the frequency of user jump behaviors and the time interval data across paths. The time interval is extracted from the comparison information of the path start time and end time. The frequency of jump behavior is obtained by the statistics of the number of cross-path jump records. Finally, the behavioral characteristics of users in multi-path access are sorted and classified to generate multi-path behavior linkage analysis data.

[0040] The user behavior prediction model building submodule calculates the conversion probability of behavior nodes based on multi-path behavior linkage analysis data, and builds a user transaction behavior prediction model based on the behavior chain characteristics from browsing to purchase and the user's behavior on the access path; To calculate the conversion probability of a behavior node, the formula is used: ; Get the slave behavior node Convert to Node The conversion probability ; in, It is a slave node Convert to Node The total number of times is extracted through the continuous node conversion records in the user behavior log, such as recording the sequence of each user's page clicks, jumps, etc., to determine the arrive The number of conversion actions, Is a node The average dwell time of users at a node The total dwell time divided by the node The number of visits is obtained, and the stay time is extracted from the timestamp data of entering and leaving the node. Is a node The total number of visits is calculated by counting the nodes in the user behavior log. For example, to record the number of times a user visits a page and mark it as a node , calculate the total number of times, Is a node The average dwell time of users at a node The total dwell time divided by the node The number of visits is obtained, and the residence time is extracted from the timestamp data of entering and leaving the node.

[0041] For example, a user has the following behavior nodes on an e-commerce platform: browsing a page , click on product details , record display node Convert to Node The total number of nodes is 120. The total number of visits is 200, and the node and nodes The total residence time is 600 seconds and 240 seconds respectively: Get Node and nodes Average length of stay: ; Calculate the conversion probability according to the formula: ; Conversion probability It shows that 40% of page browsing operations are successfully converted into clicks on product details. Combined with the time residence characteristics, it shows that the node It is more attractive to users. The historical value analysis shows that the previous conversion probability was 0.3. The 10% increase reflects the enhanced conversion efficiency after path optimization. The user's behavior chain feature data (such as conversion probability , Node residence time ) and path behavior data (such as access path length, jump frequency, etc.) to form a multi-dimensional feature table of user behavior paths, analyze the integrated feature data, and select the most relevant features for purchasing behavior prediction as model input variables, such as conversion probability It is a strongly correlated feature, while the path length may have a smaller impact. It can be combined with feature importance analysis tools (such as decision trees and random forests) for screening, and a model suitable for predictive classification problems, such as logistic regression, support vector machines, or neural networks, is used. The output result is the probability of whether the user completes the purchase behavior, with a value range of [0,1]. The historical user behavior data is used as the training set, where the label value is whether the purchase behavior is completed (1 means purchase, 0 means no purchase), and the model parameters are trained. The historical behavior chain features (such as ) and path behavior (such as jump frequency, path length), example input features: , evaluate the model performance on the validation set, use indicators such as accuracy, recall, and F1 value, adjust the model structure and parameters to optimize the prediction performance, and finally build a user transaction behavior prediction model.

[0042] See also Figure 2 and Figure 8 , e-commerce marketing data optimization modules include: The key behavior node classification submodule classifies the user's operation tendency at the key behavior nodes based on the user transaction behavior prediction model, extracts the user behavior data of completed transaction operations and the user behavior characteristics of uncompleted transaction operations in the key path, and generates key node user behavior data; Based on the user transaction behavior prediction model, the user's operation tendencies at key behavior nodes are classified. First, the user's operation data at the key nodes of the transaction behavior are extracted, such as the click operation records on the platform page, the category and quantity of goods added to the shopping cart, the length of time spent on viewing the product details page, etc. These data are sorted out through user behavior logs and classified according to whether the user has completed the transaction. The user operation behavior characteristics of completed transactions are classified as successful transaction behavior data, such as the full path characteristics from clicking on the product to the final payment. At the same time, the operation behavior characteristics of users who have not completed transactions are extracted, such as the situation where the user stays on the product page for a long time but has not completed the payment, to form an unfinished transaction behavior feature data set. The two types of data are structured and sorted, and the user behavior characteristics of successful transactions and unfinished transactions are respectively generated into key behavior feature tables. Finally, the two types of feature tables are combined to integrate and generate key node user behavior data.

[0043] The path node impact analysis submodule analyzes the impact of path nodes on user behavior chains based on key node user behavior data, cross-analyzes the correlation characteristics of abnormal behaviors and unfinished transaction paths, and generates path node impact analysis data; Based on the user behavior data of key nodes, the degree of influence of path nodes on the user behavior chain is analyzed. By calculating the number of visits, average stay time and conversion rate of each path node, the operation log information of users at the path nodes is extracted, such as the number of clicks, page loading time, the proportion of jumping to the next node, etc., and the abnormal behavior data is combined to analyze the characteristics of the path nodes that may affect the user behavior chain. The paths of abnormal behavior users are compared with the paths of completed transaction users. For example, compare which nodes the users who have not completed the transaction have stayed for a long time but have not jumped to the next step, mark these nodes as nodes with possible problems, further analyze the operation difficulty, page layout complexity, loading speed and other characteristics of these nodes, and combine with the characteristic data of the unfinished transaction path, such as the user's operation frequency and jump direction when the path is interrupted, identify the correlation characteristics between the path node and the user behavior interruption, and finally generate the path node impact analysis data.

[0044] The marketing behavior chain adjustment submodule adjusts the behavior chain data structure based on the path node impact analysis data, optimizes the node layout of the key path, and generates marketing behavior chain optimization data.

[0045] Based on the path node impact analysis data, adjust the behavior chain data structure. First, optimize the node design in the key path and arrange the nodes with high conversion rates in advance. For example, move the button for users to complete payment on the product details page from the bottom to a prominent area. At the same time, for nodes with high user churn rates, analyze their specific problems and design optimization plans. For example, reduce the number of options on complex multi-option pages and optimize page loading speed. Combined with the user's behavior data in the path, for example, add guidance prompts to nodes with low jump frequency, such as prompting the use of coupons on the checkout page. For the situation where the node jump frequency in the path is high but the payment is not completed, optimize the jump logic by adjusting the node order in the path. Finally, rebuild the optimized marketing behavior chain data structure and generate marketing behavior chain optimization data.

[0046] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0047] In the present invention, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0048] It should be understood that in various embodiments of the present invention, 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 invention.

[0049] Those skilled 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 the present invention.

[0050] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0051] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device 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 device, 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 through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0052] 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.

[0053] In addition, each functional unit in each embodiment of the present invention 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.

[0054] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

[0055] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. An AI-based e-commerce marketing optimization system, characterized by: The system comprises: The user behavior analysis module collects the user's behavior data on the e-commerce platform, divides the user's behavior data on the e-commerce platform into multiple behavior categories, and forms user behavior statistics; The cross-domain behavior analysis module analyzes the user's cross-platform behavior path based on the user behavior statistics, selects the priority access path, analyzes the active time period and path of the user's behavior on the platform, and generates the user's cross-platform path distribution data; The abnormal order identification module obtains the operation records of the order payment status in the user behavior statistics, marks the users who cancel the orders for multiple times in a row, and obtains the abnormal order data analysis results; The user stratification and dynamic feature extraction module identifies the change trend of order cancellation behavior in each access category target time period based on the user's cross-platform path distribution data and abnormal order data analysis results, and generates a stratified user feature extraction result; The user transaction pattern analysis module analyzes the user's behavior chain on the access path based on the hierarchical user feature extraction result, determines the stability of the behavior chain within the path, and builds a user transaction behavior prediction model; The e-commerce marketing data optimization module classifies the user's future operation tendencies at key behavior nodes through the user transaction behavior prediction model, cross-analyzes the associated features of abnormal behaviors and unfinished transaction paths, adjusts the behavior chain data structure, and generates marketing behavior chain optimization data.

2. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The user behavior statistics include statistics on the user's page browsing time, analysis results on product click-through rates, and distribution of types and quantities of products added to the shopping cart; the user cross-platform path distribution data include the cumulative results of the number of user visits to the platform, path classification statistics on the number of behavior jumps, and ranking results of the proportion of path visits; the abnormal order data analysis results include the distribution of each user's order cancellation rate, a list of marked users who have continuously canceled orders, and order records associated with canceled orders; the stratified user feature extraction results include user distribution statistics classified by single-path access and multi-path access, data analysis of the concentration of priority access paths, and distribution trends of users in active time periods; the user transaction behavior prediction model includes analysis results of the node stay time in the single-path access user behavior chain and stability data of the behavior sequence; the marketing behavior chain optimization data include user behavior feature comparison analysis results, evaluation data on the degree of influence of path nodes on the behavior chain, and user behavior chain optimization after critical path adjustment.

3. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The user behavior analysis module includes: The user behavior data collection submodule collects the user's page browsing time, product click-through rate, type and quantity of products added to the shopping cart, order payment status and product access frequency based on the user's behavior data on the e-commerce platform, and generates the original user behavior data; The user behavior classification and sorting submodule divides the user behavior into multiple categories such as browsing, selecting, and purchasing based on the original user behavior data, and sorts them by platform and user behavior type to generate user behavior classification data; The user behavior statistics submodule counts the key indicators of each category of user behavior based on the user behavior classification data, including the total browsing behavior duration, selection behavior frequency, and order payment ratio of purchase behavior, and generates user behavior statistics results.

4. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The cross-domain behavior analysis module includes: The cross-platform behavior extraction submodule extracts the number of user visits and behavior jumps between multiple platforms based on the user behavior statistics, analyzes the jump frequency of the behavior path, and generates user cross-platform access path data; The path frequency statistics submodule summarizes the behavior jump frequency according to the path based on the user's cross-platform access path data, obtains the visit frequency ratio of each path, sorts the path visit frequency, and generates path access frequency data; The user active distribution analysis submodule analyzes the user's active time periods and path characteristics across multiple platforms based on the path access frequency data and combined with the classified statistics of the time periods in which the user's behavior occurs, and generates user cross-platform path distribution data.

5. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The abnormal order identification module includes: The order record processing submodule extracts the payment success record and payment cancellation record based on the operation record of the order payment status in the user behavior statistics, counts the number of order submissions and cancellations for each user, and generates user order record statistics; The order cancellation rule analysis submodule obtains the order cancellation rate of each user based on the statistical data of the user order records, counts the number and proportion of users within the cancellation rate range, analyzes the distribution of the order cancellation rate in the user group, and generates an order cancellation rate analysis result; The abnormal order marking submodule marks users who have canceled orders multiple times in a row based on the order cancellation rate analysis results, extracts order records associated with the cancellation behavior, and generates abnormal order data analysis results.

6. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The user stratification and dynamic feature extraction module includes: The user behavior path stratification submodule stratifies the user behavior paths according to the concentration of the priority access paths and the distribution of users in the active time period based on the user cross-platform path distribution data, and generates user behavior path stratification data; The path cancellation rate analysis submodule generates path cancellation rate distribution data based on the analysis results of the user behavior path hierarchical data and abnormal order data and the path access category, and calculates the order cancellation rate distribution of single-path access and multi-path access users; The behavior trend identification submodule analyzes the dynamic change characteristics of the order cancellation behavior based on the path cancellation rate distribution data and the change trend of the order cancellation behavior in the target time period, and generates a hierarchical user feature extraction result.

7. The AI-based e-commerce marketing optimization system according to claim 6, characterized in that: To analyze the dynamic characteristics of order cancellation behavior, according to the formula: ； Calculation time period Weighted average cancellation rate ; in, Is a user In time period The order cancellation rate Is a user The path complexity weight of It is the total number of all users on the e-commerce platform.

8. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The user transaction pattern analysis module includes: The single-path behavior chain analysis submodule analyzes the behavior path of the single-path access user based on the hierarchical user feature extraction result, identifies the sequence and stay characteristics of the behavior nodes, determines the stability of the behavior chain, and generates single-path behavior chain analysis data; The multi-path behavior linkage analysis submodule analyzes the cross-path behavior distribution of multi-path access users based on the single-path behavior chain analysis data, determines the behavior linkage characteristics between associated paths, and generates multi-path behavior linkage analysis data; The user behavior prediction model building submodule calculates the conversion probability of the behavior node based on the multi-path behavior linkage analysis data, and builds a user transaction behavior prediction model according to the behavior chain characteristics from browsing to purchasing and the user's behavior on the access path.

9. The AI-based e-commerce marketing optimization system according to claim 8, characterized in that: To calculate the conversion probability of a behavior node, the formula is used: ； Get the slave behavior node Convert to Node The conversion probability ; in, It is a slave node Convert to Node The total number of times, Is a node The average length of stay, Is a node Total visits, Is a node average stay time.

10. The AI-based e-commerce marketing optimization system according to claim 1, characterized in that: The e-commerce marketing data optimization module includes: The key behavior node classification submodule classifies the user's operation tendency at the key behavior nodes based on the user transaction behavior prediction model, extracts the user behavior data of completed transaction operations and the user behavior characteristics of uncompleted transaction operations in the key path, and generates key node user behavior data; The path node impact analysis submodule analyzes the impact of the path node on the user behavior chain based on the key node user behavior data, cross-analyzes the correlation characteristics of abnormal behavior and unfinished transaction paths, and generates path node impact analysis data; The marketing behavior chain adjustment submodule adjusts the behavior chain data structure based on the path node impact analysis data, optimizes the node layout of the key path, and generates marketing behavior chain optimization data.

Images