Shopping system based on WeChat applet and use method

Abstract

The invention relates to the technical field of Internet shopping, in particular to a shopping system based on a WeChat applet and a use method, and the method comprises the steps: obtaining shopping cart operation behavior sequence data through WeChat applet front-end event monitoring; extracting time dimension and space dimension features, and outputting a decision vector of a three-dimensional orthogonal component through a vector synthesis algorithm and orthogonal normalization; establishing a strategy template library, calculating an Euclidean distance between a decision vector and a template to match a strategy, and adjusting an intensity parameter according to the vector modulus length; according to the method, user behaviors are converted into three-dimensional orthogonal vector purchase intention intensity, price sensitivity and decision hesitance, each component is calculated through algorithms such as logarithmic growth rate, maximum time interval and standard deviation, numerical expression of behavior characteristics is achieved, limitation of a traditional binarization model is broken through, and the user experience is improved. Discrete operation behaviors are quantified into computable and comparable feature vectors, so that merchants can accurately analyze user decision tendencies from multiple dimensions.

Description

Technical Field

[0001] The present invention relates to the technical field of Internet shopping, and in particular to a shopping system based on WeChat mini-programs and a method for using the same. Background Art

[0002] With the rapid development of mobile Internet technology and the popularization of smart phones, the scale of the global e-commerce market continues to expand. Data shows that the scale of Internet users is climbing steadily, and the proportion of mobile transactions in the total online retail sales continues to rise. Users' demand for convenient, efficient and instant shopping experiences is growing. Against this background, e-commerce is gradually shifting from traditional PC to mobile, driving the evolution of e-commerce platforms towards intelligence and scenario-based development. Against this background, WeChat mini-programs, with their "no need to download and install, just use it and go" feature, have become an important carrier connecting users and services. Users do not need to download and install, and can reach them in seconds through scanning codes, searching, etc., which is in line with fragmented consumption habits.

[0003] The existing WeChat mini-program shopping system can only record static operation results such as adding and deleting products, and cannot capture the temporal characteristics of user operations, resulting in an inability to deeply analyze the user decision-making process. It relies on simple binary judgments and lacks quantitative analysis of multi-dimensional characteristics such as user purchase intention intensity, price sensitivity, and decision hesitation, making it difficult to accurately characterize user needs. Summary of the Invention

[0004] The present invention addresses the technical problems existing in the prior art and provides a shopping system based on WeChat mini-programs and a method for using the same.

[0005] The technical solution of the present invention to solve the above technical problems is as follows: A shopping system based on WeChat mini-program, comprising: Operation behavior sequence collection module: obtains shopping cart operation behavior sequence data through WeChat applet front-end event monitoring; 3D orthogonal component decision vector generation module: extracts time and space dimension features, and outputs the 3D orthogonal component decision vector through vector synthesis algorithm and orthogonal normalization; Settlement guidance strategy trigger module: establish a strategy template library, calculate the Euclidean distance between the decision vector and the template to match the strategy, and adjust the strength parameter according to the vector modulus.

[0006] In a preferred embodiment, the operation behavior sequence acquisition module obtains the shopping cart operation behavior sequence in the WeChat applet. The operation behavior sequence includes three core data: the number of views, the number of times the user has added the item after deleting it, and the comparison operation duration. The number of views is obtained by capturing the user's touch behavior through the applet's native event monitoring mechanism, and the screen coordinates, focus dwell time, and associated product ID of each operation point are simultaneously recorded. During the acquisition process, double validity filtering is enabled, and only valid operation points with a screen movement speed lower than a first speed threshold are retained. The screening of the valid operation points is performed by the coordinates of the i-th operation point ( ), timestamp , calculate the moving speed of adjacent operating points, moving speed The specific calculation formula is as follows:

[0007] in, Indicates the timestamp of the i-th operation point, Indicates the screen horizontal coordinate of the i-1th operating point, Indicates the screen vertical coordinate of the i-1th operating point, Indicates the timestamp of the i-1th operation point. Valid operation points must satisfy ,in, Represents the first speed threshold. An operation point that meets the conditions is a valid operation point. Valid operation points that meet the conditions are counted to obtain the number of valid operation points. At the same time, the native event monitoring unit is used to ensure timing consistency. The number of views is accumulated based on the product ID dimension. When a user clicks on a product card, the count is automatically incremented. A memory period is set for deletion operations. If the same product ID is added again within the memory period, the number of re-additions after deletion is automatically incremented. Based on the spatial distribution characteristics within the preset dynamic analysis window, a density clustering algorithm with a specific neighborhood radius and minimum number of points is used to cluster the effective operating points. For an operating point p, its neighborhood is defined as:

[0008] in, = is the pixel coordinate distance, represents the neighborhood radius, represents the operating point p Neighborhood point set, that is, the distance from p is no more than All points of represents the Euclidean distance between the operating points p and q, represents the neighborhood radius, p and q represent the coordinate variables of any two operating points, if | |≥MinPts, then p is the core point. Starting from the core point, all points with reachable density are connected to form a cluster, and the formed cluster is used as the convex hull. Among them, MinPts represents the minimum point number threshold. After calculating the convex hull boundary of each cluster, the interface hot zone boundary is generated by the curve smoothing algorithm. This process requires that each window contains no less than the minimum number of valid operation points. At the same time, a noise filtering mechanism is implemented to eliminate short-term operation points with a residence time less than the first time threshold and abnormal points whose adjacent point velocity changes exceed the acceleration threshold. The specific calculation formula is as follows:

[0009] in, It represents the speed change rate between the ith operating point and the previous point, and eliminates those that satisfy | |> points, where It represents the acceleration threshold, which is used to eliminate abnormal points with sudden speed changes. When the density of valid points is lower than a specific density threshold, data compensation is performed on the virtual operating points generated based on the historical heat map in a proportional weighted manner. Assuming the screen area is S and the number of valid points is n, the density is: ,when Compensation is triggered when Indicates the density threshold and stores the behavior data with spatiotemporal labels in a structured format, including timestamps, screen coordinates, and dwell time. The number of times a user adds a new item after deletion is obtained by automatically associating the hot zone boundary with the product ID. If the hot zone covers the product card area and the dwell time exceeds the first time threshold, the operation point is marked as a valid viewing behavior. For comparison operations, the long press event of the multiple-select button is monitored, starting from the time the first item is checked and ending when the item is unchecked and the comparison is submitted. A comparison operation duration record is generated. All behavior data is sorted by timestamp and stored in the local cache. The number of views and the number of times a user adds a new item after deletion are accumulated using the product ID as the index. The duration of the comparison operation is obtained by storing the complete time series record of the comparison operation event with the session ID as the index. When the density of valid points is insufficient, the virtual operation points generated based on the historical heat map will be assigned an associated ID according to the historical click-through rate of the product.

[0010] In a preferred embodiment, the three-dimensional orthogonal component decision vector generation module extracts features of the time dimension and the space dimension. The time dimension calculates the operation frequency per unit time by counting the timestamp distribution density of valid operation points within a preset dynamic analysis window. The specific calculation formula is as follows:

[0011] Among them, f represents the operation frequency per unit time, which represents the activeness of user behavior. Indicates the number of valid operating points within the preset dynamic analysis window, Indicates the duration of the preset dynamic analysis window, which is used to characterize the activity of user behavior, analyze the discreteness of the operation point timestamps, and reflect the persistence of the behavior by calculating the timestamp standard deviation. The specific calculation formula for the timestamp standard deviation is as follows:

[0012] in, It represents the standard deviation of the timestamp, reflecting the discrete degree of the time distribution of the operation points. The larger the value, the stronger the suddenness of the behavior. m represents the number of valid operation points in the window. Indicates the timestamp of the i-th valid operation point, It represents the average value of the timestamps of all valid operation points in the window. The spatial dimension is calculated by calculating the overlapping area ratio of the hot zones corresponding to different products. If the overlapping area exceeds the preset threshold, it is determined to be a strongly associated product. The specific calculation formula for the overlapping area ratio of the hot zones is as follows:

[0013] in, Indicates the overlapping area ratio of the hot zones corresponding to two products, which is used to judge the strength of product association. The overlapping area of ​​the two hot zones, Respectively represent the area of ​​the hot zones corresponding to the two commodities, The smaller of the two hotspot areas is used as the normalization benchmark to avoid the impact of area differences on the proportion. The number of times users jump between different hotspots is also counted to form a spatial jump frequency matrix for product operations. The features of the time dimension and the space dimension are input into the preset vector synthesis algorithm, and the output is a decision vector containing three orthogonal components, including: The first orthogonal component represents the purchase intention intensity: The number of views within the window is accumulated by product ID to obtain the density of views per unit time. The specific calculation formula is as follows:

[0014] in, represents the viewing density, Indicates the cumulative number of times a product is viewed within a preset window. It calculates the average length of time users spend in a product's hot zone. The specific calculation formula is as follows:

[0015] in, Indicates the average duration of stay in the hot zone, represents the length of time the i-th effective operating point stays in the hot zone, n represents the number of effective operating points in the hot zone, and finally the concentration is measured by the distribution variance of the hot zone operating points. The specific calculation formula for the operating point concentration is as follows:

[0016] in, represents the concentration of operating points, ( , ) represents the screen coordinates of the i-th valid operating point, ( , ) represents the average value of the hot zone operating point coordinates. After standardizing the above three indicators, the specific calculation formula for standardization is as follows:

[0017] in, , , represents the normalized index, , The historical mean and standard deviation of the corresponding indicator are weighted and summed according to the preset weights to obtain the purchase intention intensity value. The specific calculation formula of the weighted summation is as follows: +

[0018] in, Indicates the purchase intention strength value, , , Indicates the preset weight coefficient; The second orthogonal component characterizes price sensitivity: The cumulative number of times the same product was deleted and then added again during the memory period was counted. The time interval between each re-addition and deletion was calculated, and the maximum time interval was taken as a reference. The relationship between the number of times and the time interval was fitted using a logistic regression model to output the price sensitivity value. The specific calculation formula for the price sensitivity value is as follows:

[0019] in, Indicates the price sensitivity value, ranging from (0,1). Indicates the cumulative number of times the same product is deleted and then added again during the memory period. Indicates the maximum time interval between re-addition and deletion during the memory period. 、 、 represents the logistic regression model parameters; The third orthogonal component represents the decision hesitation: Get the duration of the comparison operation, and then calculate the standard deviation of the duration. The specific calculation formula is as follows:

[0020] in, Indicates the standard deviation of the comparison operation time, reflecting time fluctuations, Indicates the duration of the i-th comparison operation, represents the average duration of the comparison operation, k represents the number of comparison operations, and the standard deviation reflects the fluctuation of the user's time in different comparison sessions. If there are multiple comparison operations, the number of compared products and the number of hot zone jumps are further analyzed. The entropy weight method is used to calculate the weight of each indicator and the decision hesitation value is obtained comprehensively. The specific calculation formula is as follows:

[0021] in, Indicates the decision hesitation, ranging from [0,1]. For example, the longer the comparison time and the greater the fluctuation, the higher the hesitation; The decision vector of the three orthogonal components synthesized into three-dimensional orthogonal components in orthogonal directions Normalization is performed to make the vector modulus in the range [0, 1]. At the same time, a hot zone density attenuation factor is introduced. If the density of valid points in the current window is lower than the density threshold, the vector component needs to be multiplied by the attenuation coefficient to correct the impact of data compensation on the vector accuracy.

[0022] In a preferred embodiment, the settlement guidance policy triggering module first constructs a policy template library, stores three basic policy templates in a cloud database, including: Template A for issuing discount coupons, Template B for prompting low inventory, and Template C for recommending similar products, and calculates the Euclidean distance between the decision vector of the three-dimensional orthogonal component and each template. The specific calculation formula is as follows:

[0023] in, The Euclidean distance between the decision vector representing the three-dimensional orthogonal component and the k-th strategy template is used to measure the degree of matching. The smaller the value, the higher the matching degree. , , Indicates the preset component threshold of the kth strategy template, such as the price sensitivity threshold of template A, the purchase intention intensity threshold of template B, etc. k represents the strategy template number, The template with the smallest distance is selected as the basic strategy, and the strategy parameters are adjusted according to the vector modulus. If the distances of multiple templates are all less than the threshold, they are weighted and combined according to the inverse of the distance, and the main strategy is displayed first.

[0024] An embodiment of the present invention further provides a method for using a shopping system based on a WeChat applet, comprising the following steps: S101. Obtain shopping cart operation behavior sequence data through WeChat mini-program front-end event monitoring; S102, extracting time dimension and space dimension features, and outputting a decision vector of three-dimensional orthogonal components through a vector synthesis algorithm and orthogonal normalization; S103: Establish a policy template library, calculate the Euclidean distance between the decision vector and the template to match the policy, and adjust the strength parameter according to the vector modulus.

[0025] The beneficial effects of the present invention are as follows: the present invention uses the WeChat applet front-end event monitoring mechanism to collect core data such as the number of times users view shopping cart items, the number of times they delete and re-add items, and the comparison of operation duration in real time. Each operation is accompanied by a timestamp, and a density clustering algorithm is used to construct a spatiotemporal map of user behavior, converting user behavior into three-dimensional orthogonal vectors of purchase intention intensity, price sensitivity, and decision hesitation. Each component is calculated through algorithms such as logarithmic growth rate, maximum time interval, and standard deviation to achieve numerical expression of behavioral characteristics, breaking through the limitations of traditional binary models, and quantifying discrete operation behaviors into computable and comparable feature vectors, enabling merchants to accurately analyze user decision-making tendencies from multiple dimensions, a dynamic strategy triggering mechanism, and a personalized operation closed loop. Three types of strategy templates are preset for discount coupon issuance, inventory emergency prompts, and similar product recommendations. By calculating the Euclidean distance between the decision vector and the template, the optimal strategy is automatically matched, and the strategy strength is dynamically adjusted according to the vector modulus. At the same time, the strategy template is continuously optimized through user feedback data to form a self-iterative intelligent operation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the method of the present invention; Figure 2 is a system block diagram of the present invention; Figure 3 This is a logic block diagram of the three-dimensional orthogonal component decision vector generation module of the present invention. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0028] In the description of the present application, the terms "first", "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "a plurality of" is two or more, unless otherwise explicitly specified.

[0029] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or description". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that those skilled in the art can realize the present application without using these specific details. In other examples, well-known structures and processes will not be described in detail to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope in accordance with the principles and characteristics disclosed in the present application.

[0030] As Figure 1 The embodiment provides a use method of a shopping system based on a WeChat applet, including the following steps: S101, acquiring shopping cart operation behavior sequence data through WeChat applet front-end event listening; S102, extracting time dimension and space dimension features, outputting three-dimensional orthogonal component decision vectors through vector synthesis algorithm and orthogonal normalization; S103, establishing a strategy template library, calculating the Euclidean distance between the decision vector and the template to match the strategy, and adjusting the intensity parameter according to the vector module length.

[0031] As Figure 2 The embodiment provides a shopping system based on a WeChat applet, including: The operation behavior sequence acquisition module acquires shopping cart operation behavior sequence data through WeChat applet front-end event listening; In this embodiment, it is necessary to specifically explain the operation behavior sequence acquisition module. The operation behavior sequence acquisition module obtains the shopping cart operation behavior sequence in the WeChat applet. The operation behavior sequence includes three core data: the number of views, the number of times of re-addition after deletion, and the comparison operation duration. The acquisition of the number of views captures the user's touch behavior through the applet's native event monitoring mechanism, and simultaneously records the screen coordinates, focus dwell time, and associated product ID of each operation point. During the acquisition process, double validity filtering is enabled, and only valid operation points with a screen movement speed below the first speed threshold are retained. The screening of the valid operation points is based on the coordinates of the i-th operation point ( ), timestamp , calculate the moving speed of adjacent operating points, moving speed The specific calculation formula is as follows:

[0032] in, Indicates the timestamp of the i-th operation point, Indicates the screen horizontal coordinate of the i-1th operating point, Indicates the screen vertical coordinate of the i-1th operating point, Indicates the timestamp of the i-1th operation point. Valid operation points must satisfy ,in, Represents the first speed threshold. An operation point that meets the conditions is a valid operation point. Valid operation points that meet the conditions are counted to obtain the number of valid operation points. At the same time, the native event monitoring unit is used to ensure timing consistency. The number of views is accumulated based on the product ID dimension. When a user clicks on a product card, the count is automatically incremented. A memory period is set for deletion operations. If the same product ID is added again within the memory period, the number of re-additions after deletion is automatically incremented. Based on the spatial distribution characteristics within the preset dynamic analysis window, a density clustering algorithm with a specific neighborhood radius and minimum number of points is used to cluster the effective operating points. For an operating point p, its neighborhood is defined as:

[0033] in, = is the pixel coordinate distance, represents the neighborhood radius, represents the operating point p Neighborhood point set, that is, the distance from p is no more than All points of represents the Euclidean distance between the operating points p and q, represents the neighborhood radius, p and q represent the coordinate variables of any two operating points, if | |≥MinPts, then p is the core point. Starting from the core point, all points with reachable density are connected to form a cluster, and the formed cluster is used as the convex hull. Among them, MinPts represents the minimum point number threshold. After calculating the convex hull boundary of each cluster, the interface hot zone boundary is generated by the curve smoothing algorithm. This process requires that each window contains no less than the minimum number of valid operation points. At the same time, a noise filtering mechanism is implemented to eliminate short-term operation points with a residence time less than the first time threshold and abnormal points whose adjacent point velocity changes exceed the acceleration threshold. The specific calculation formula is as follows:

[0034] in, It represents the speed change rate between the ith operating point and the previous point, and eliminates those that satisfy | |> points, where It represents the acceleration threshold, which is used to eliminate abnormal points with sudden speed changes. When the density of valid points is lower than a specific density threshold, data compensation is performed on the virtual operating points generated based on the historical heat map in a proportional weighted manner. Assuming the screen area is S and the number of valid points is n, the density is: ,when Compensation is triggered when Indicates the density threshold and stores the behavior data with spatiotemporal labels in a structured format, including timestamps, screen coordinates, and dwell time. The number of times a user adds a new item after deletion is obtained by automatically associating the hot zone boundary with the product ID. If the hot zone covers the product card area and the dwell time exceeds the first time threshold, the operation point is marked as a valid viewing behavior. For comparison operations, the long press event of the multiple-select button is monitored, starting from the time the first item is checked and ending when the item is unchecked and the comparison is submitted. A comparison operation duration record is generated. All behavior data is sorted by timestamp and stored in the local cache. The number of views and the number of times a user adds a new item after deletion are accumulated using the product ID as the index. The duration of the comparison operation is obtained by storing the complete time series record of the comparison operation event with the session ID as the index. When the density of valid points is insufficient, the virtual operation points generated based on the historical heat map will be assigned an associated ID according to the historical click-through rate of the product.

[0035] It should be noted that the comparison operation duration refers to the duration of the user's continuous operations such as comparing product parameters and weighing prices after selecting multiple items in the shopping cart and entering the comparison mode. Specifically: Trigger condition: When the user selects at least two items at the same time (such as clicking a multiple-select button, long pressing a product card, etc.), the system will enter comparison mode and the timer will start at this time.

[0036] Time calculation: The comparison operation duration starts from the moment the product is first selected and ends when the user cancels the comparison, submits the order, or leaves the comparison interface.

[0037] Behavioral significance: This duration reflects the user's hesitation in making decisions between different products. A longer duration usually means that the user has made a deeper trade-off in terms of price, performance, and reviews, and is more cautious in the decision-making process. Conversely, a shorter duration may indicate that the user has a clear goal or makes a quick choice after comparison.

[0038] It should be noted that the interface hot zone refers to the high-frequency operation area generated by algorithm clustering and boundary processing based on the distribution density of user operation behaviors in the interactive interface. It is essentially a visual presentation of user operation habits.

[0039] It should be noted that the specific steps for generating interface hot zone coordinates are as follows: S1. Calculate the convex hull boundary: Use the minimum convex polygon to wrap the clustered operation point set to quickly define the outer boundary of the operation-intensive area. For example, when the operation point distribution is irregular, the convex hull can extract internal redundant points and retain only the outermost contour points to prevent the hot zone from excessively including non-core operation areas. The convex hull follows the minimum enclosing principle to ensure that all operation points are included in the hot zone and avoid missing edge operation points. For example, when the user frequently touches the edge of the screen, the convex hull can accurately capture these boundary points to prevent the hot zone from shrinking. Directed distance calculation formula: d

[0040] Calculate the directed distance from point p to baseline AB to determine whether the point is on the left or right side of the baseline. A positive value indicates that the point is on the left side of the baseline, a negative value indicates that the point is on the right side, and 0 indicates collinearity. The point set is divided into and , repeatedly calculate the directed distance until all points are included in the convex hull boundary, recursively find the point farthest from the current baseline, divide the point set into sub-problems, and gradually construct the convex hull boundary to ensure that the convex hull contains all peripheral points; S2. Quadratic B-spline curve smoothing calculation: Quadratic B-spline curve smoothing calculation eliminates convex hull corners and fits the operation distribution. The polygonal boundary generated by the convex hull has sharp corners, while the hot zone boundary of the actual user operation is closer to the natural transition of the smooth curve such as the finger sliding trajectory. The quadratic B-spline is calculated by weighted average of adjacent vertices ( , , ), converting the edges and corners into smooth curves, which is more in line with the gradual characteristics of the operation density. The quadratic B-spline only needs 3 adjacent points to calculate the intermediate smooth point ( ), the quadratic B-spline avoids overfitting while retaining the basic shape characteristics of the convex hull; The middle vertex of the convex hull , its smooth point From the previous vertex , current vertex , the next vertex By weight , , Weighted average generation to achieve smooth curves and smooth points It is generated by weighting three adjacent vertices. The specific calculation formula is as follows:

[0041] Expand the coordinates of three adjacent vertices ,

[0042] Convex hull first endpoint and tail endpoint and adjacent points by weight respectively , Weighting is used to avoid endpoint mutations and ensure the continuity of the curve. The specific calculation formula is as follows: First endpoint Smooth points:

[0043] Tail endpoint Smooth points:

[0044] The smoothed hot zone boundary point set is C= ,in A continuous smooth curve.

[0045] The convex hull serves as the initial boundary, providing a rough geometric framework for the hotspot. B-spline smoothing then optimizes the details, making the hotspot more consistent with the natural distribution of user operations. Ultimately, the hotspot coordinates on the interface are generated to be both complete and smooth. S3. Hot zone coordinate output formula: Calculate the extreme coordinates of C:

[0046] The rectangular hot zone is represented by ( , , , ), use the smooth point set C as the polygon vertices and connect them in sequence , forming a closed polygonal boundary.

[0047] 3D orthogonal component decision vector generation module: extracts time and space dimension features, and outputs the 3D orthogonal component decision vector through vector synthesis algorithm and orthogonal normalization; In this embodiment, a three-dimensional orthogonal component decision vector generation module needs to be specifically explained. The three-dimensional orthogonal component decision vector generation module extracts features of the time dimension and the space dimension. The time dimension calculates the operation frequency per unit time by counting the timestamp distribution density of valid operation points within a preset dynamic analysis window (e.g., 30 minutes). The specific calculation formula is as follows:

[0048] Among them, f represents the operation frequency per unit time, which represents the activeness of user behavior. Indicates the number of valid operating points within the preset dynamic analysis window, Indicates the duration of the preset dynamic analysis window, which is used to characterize the activity of user behavior, analyze the discreteness of the operation point timestamps, and reflect the persistence of the behavior by calculating the timestamp standard deviation. The specific calculation formula for the timestamp standard deviation is as follows:

[0049] in, It represents the standard deviation of the timestamp, reflecting the discrete degree of the time distribution of the operation points. The larger the value, the stronger the suddenness of the behavior. m represents the number of valid operation points in the window. Indicates the timestamp of the i-th valid operation point, It represents the average value of the timestamps of all valid operation points in the window. The spatial dimension is calculated by calculating the overlapping area ratio of the hot zones corresponding to different products. If the overlapping area exceeds the preset threshold, it is determined to be a strongly associated product. The specific calculation formula for the overlapping area ratio of the hot zones is as follows:

[0050] in, Indicates the overlapping area ratio of the hot zones corresponding to two products, which is used to judge the strength of product association. The overlapping area of ​​the two hot zones, Respectively represent the area of ​​the hot zones corresponding to the two commodities, The smaller of the two hotspot areas is used as the normalization benchmark to avoid the impact of area differences on the proportion. The number of times users jump between different hotspots is also counted to form a spatial jump frequency matrix for product operations. The features of the time dimension and the space dimension are input into the preset vector synthesis algorithm, and the output is a decision vector containing three orthogonal components, including: The first orthogonal component represents the purchase intention intensity: The core indicator is the view count density, which is calculated by combining the hot zone dwell time and the operation point concentration. Specifically, the view counts within the window are first accumulated by product ID to obtain the view count density per unit time. The specific calculation formula is as follows:

[0051] in, represents the viewing density, Indicates the cumulative number of views of a product within the preset window. Secondly, the average length of time users stay in the product hot zone is calculated, that is, the sum of the length of time spent at all valid operation points in the hot zone / the number of valid points in the hot zone. The specific calculation formula is as follows:

[0052] in, Indicates the average duration of stay in the hot zone, represents the length of time the i-th effective operating point stays in the hot zone, n represents the number of effective operating points in the hot zone, and finally the concentration is measured by the distribution variance of the hot zone operating points. The smaller the variance, the higher the concentration. The specific calculation formula for the operating point concentration is as follows:

[0053] in, represents the concentration of operating points, ( , ) represents the screen coordinates of the i-th valid operating point, ( , ) represents the average value of the hot zone operating point coordinates. After standardizing the above three indicators, the specific calculation formula for standardization is as follows:

[0054] in, , , represents the normalized index, , The historical mean and standard deviation of the corresponding indicator are weighted and summed according to the preset weights to obtain the purchase intention intensity value. The specific calculation formula of the weighted summation is as follows: +

[0055] in, Indicates the purchase intention strength value, , , Indicates the preset weight coefficient; The second orthogonal component characterizes price sensitivity: Focusing on the number of times a product is deleted and then added again, and the corresponding time interval, we first count the cumulative number of times the same product is deleted and then added again during the memory period. We then calculate the time interval between each re-addition and deletion, taking the maximum time interval as a reference. We then use a logistic regression model to fit the relationship between the number of times and the time interval, and output the price sensitivity value. The specific calculation formula for the price sensitivity value is as follows:

[0056] in, Indicates the price sensitivity value, ranging from (0,1). Indicates the cumulative number of times the same product is deleted and then added again during the memory period. Indicates the maximum time interval between re-addition and deletion during the memory period. 、 、 represents the logistic regression model parameters, e.g., more times of re-addition and shorter time intervals indicate higher sensitivity; The third orthogonal component represents the decision hesitation: Focusing on the comparison operation duration and its fluctuation, we first obtain the duration of the comparison operation and then calculate the standard deviation of the duration. The specific calculation formula is as follows:

[0057] in, Indicates the standard deviation of the comparison operation time, reflecting time fluctuations, Indicates the duration of the i-th comparison operation, represents the average duration of the comparison operation, k represents the number of comparison operations, and the standard deviation reflects the fluctuation of the user's time in different comparison sessions. If there are multiple comparison operations, the number of compared products and the number of hot zone jumps are further analyzed. The entropy weight method is used to calculate the weight of each indicator and the decision hesitation value is obtained comprehensively. The specific calculation formula is as follows:

[0058] in, Indicates the decision hesitation, ranging from [0,1]. For example, the longer the comparison time and the greater the fluctuation, the higher the hesitation; The decision vector of the three orthogonal components synthesized into three-dimensional orthogonal components in orthogonal directions And perform normalization processing to make the vector modulus in the range of [0, 1]. At the same time, introduce the hot zone density attenuation factor. If the density of valid points in the current window is lower than the density threshold, the vector component needs to be multiplied by the attenuation coefficient to correct the data compensation for the impact on the vector accuracy. The attenuation coefficient can be expressed as: Attenuation coefficient =

[0059] The specific steps for calculating the weight of each indicator using the entropy weight method are as follows: S1. Indicator matrix standardization: Each standardized indicator corresponds to a component of the decision vector of the three-dimensional orthogonal components. The original indicator is converted into a value in the interval [0, 1] to ensure that indicators of different dimensions can directly form the vector dimension.

[0060] in, represents the jth index value of the i-th sample, Indicates the normalized index value, the normalized comparison duration index value The first component of the decision vector that constitutes the three-dimensional orthogonal component, Indicates the maximum value of the jth indicator in all samples, indicating the maximum number of hot zone jumps of all samples, and is used to calculate the denominator of the range standardization; S2. Calculate information entropy: The smaller the information entropy, the greater the difference between the indicators among samples, and the more suitable it is as a key dimension to distinguish different decision objects. Indicators with low information entropy may be given higher weights in the decision vector of three-dimensional orthogonal components, making the vector more able to reflect the differences between samples;

[0061] in, represents the information entropy of the j-th indicator, ; S3. Calculate weights: The weights directly determine the contribution of each standardized indicator in the decision vector of the three-dimensional orthogonal components. A comprehensive indicator is generated by weighted summation, or used as a scaling factor for the dimension. A high-weight indicator vector has a larger component in the corresponding dimension, thereby affecting the position and direction of the vector in three-dimensional space.

[0062] in, represents the weight of the jth indicator, and J represents the total number of indicators; Settlement guidance strategy trigger module: establishes a strategy template library, calculates the Euclidean distance between the decision vector and the template to match the strategy, and adjusts the strength parameter according to the vector modulus; In this embodiment, the settlement guidance policy triggering module needs to be specifically explained. The settlement guidance policy triggering module first builds a policy template library and stores three types of basic policy templates in a cloud database: Template A is for issuing discount coupons, matching high price sensitivity, and the preset trigger condition is price sensitivity Preset threshold, strategy strength parameter dynamically adjusted according to vector modulus: discount coupon denomination = basic denomination Price sensitivity value Vector length, and no more than 20% of the original price of the commodity, wherein the base denomination is preset according to the commodity category profit space; Template B is a stock shortage prompt, matching high purchase willingness intensity, and the preset triggering condition is the purchase willingness intensity vector component The preset threshold value, the stock countdown display value = actual stock (1-purchase willingness intensity value), the minimum display is 1 piece, when the actual stock When the early warning stock (early warning stock = commodity daily sales 1.5) starts this strategy; Template C is a similar product recommendation, matching high decision-making hesitation, and the preset triggering condition is the decision-making hesitation vector component The preset threshold value, the recommendation logic is based on the commodity hot area association matrix: extract the top-N commodities with the highest overlap with the current commodity hot area, arrange them in descending order of overlap, and recommend the card to highlight the price difference and user evaluation of the commodity, guiding the user to quickly make a decision; The decision vector of the three-dimensional orthogonal component and the Euclidean distance of each template are calculated, and the specific calculation formula is as follows:

[0063] Among them, The Euclidean distance between the decision vector of the three-dimensional orthogonal component and the kth strategy template, used to measure the matching degree, the smaller the value, the higher the matching degree, , , The kth strategy template preset component threshold, such as the price sensitivity threshold of template A, the purchase willingness intensity threshold of template B, etc., k represents the strategy template number, Select the template with the smallest distance as the basic strategy, adjust the strategy parameters according to the vector length, for example, when the vector length is The face value of the discount coupon is increased by 20%, and the stock countdown time is shortened by 30%, if the distance of multiple templates is less than the threshold value, the distance is inversely weighted and combined, and the main strategy is preferentially displayed.

[0064] It should be noted that in the above content, the vector length is the square operation of the three orthogonal components of the three-dimensional orthogonal component decision vector, and the results are added and then squared, and the specific calculation formula is as follows:

[0065] Among them, The length of the three-dimensional decision vector, used to measure the overall strength of the vector, The first component of the three-dimensional orthogonal component decision vector, that is, the purchase willingness intensity value, The second component of the three-dimensional orthogonal component decision vector, that is, the price sensitivity value, The third component of the decision vector representing the three-dimensional orthogonal components is the decision hesitation value.

[0066] It should be noted that, in the above embodiments, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0067] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0068] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0069] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0071] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0072] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A shopping system based on WeChat applet, characterized in that: include: Operation behavior sequence collection module: obtains shopping cart operation behavior sequence data through WeChat applet front-end event monitoring; 3D orthogonal component decision vector generation module: extracts time and space dimension features, and outputs the 3D orthogonal component decision vector through vector synthesis algorithm and orthogonal normalization; Settlement guidance strategy trigger module: establish a strategy template library, calculate the Euclidean distance between the decision vector and the template to match the strategy, and adjust the strength parameter according to the vector modulus.

2. A shopping system based on WeChat applet according to claim 1, characterized in that: The operation behavior sequence acquisition module obtains the shopping cart operation behavior sequence in the WeChat applet. The operation behavior sequence includes three core data: the number of views, the number of times the user adds the item after deleting it, and the comparison operation duration. The number of views is obtained by capturing the user's touch behavior through the applet's native event monitoring mechanism, and the screen coordinates, focus dwell time, and associated product ID of each operation point are simultaneously recorded. During the acquisition process, double validity filtering is enabled to only retain valid operation points whose screen movement speed is lower than the first speed threshold. The screening of the valid operation points is based on the coordinates of the i-th operation point ( ), timestamp , calculate the moving speed of adjacent operating points, moving speed The specific calculation formula is as follows: in, Indicates the timestamp of the i-th operation point, Indicates the screen horizontal coordinate of the i-1th operating point, Indicates the screen vertical coordinate of the i-1th operating point, Indicates the timestamp of the i-1th operation point. Valid operation points must satisfy ,in, Represents the first speed threshold. An operation point that meets the conditions is a valid operation point. Valid operation points that meet the conditions are counted to obtain the number of valid operation points. At the same time, the native event monitoring unit is used to ensure timing consistency. The number of views is accumulated based on the product ID dimension. When a user clicks on a product card, the count is automatically incremented. A memory period is set for deletion operations. If the same product ID is added again within the memory period, the number of re-additions after deletion is automatically incremented. Based on the spatial distribution characteristics within the preset dynamic analysis window, a density clustering algorithm with a specific neighborhood radius and minimum number of points is used to cluster the effective operating points. For an operating point p, its neighborhood is defined as: in, = is the pixel coordinate distance, represents the neighborhood radius, represents the operating point p Neighborhood point set, that is, the distance from p is no more than All points of represents the Euclidean distance between the operating points p and q, represents the neighborhood radius, p and q represent the coordinate variables of any two operating points, if | |≥MinPts, then p is the core point. Starting from the core point, all points with reachable density are connected to form a cluster, and the formed cluster is used as the convex hull. Among them, MinPts represents the minimum point number threshold. After calculating the convex hull boundary of each cluster, the interface hot zone boundary is generated by the curve smoothing algorithm. At the same time, a noise filtering mechanism is implemented to eliminate short-term operation points with a residence time less than the first time threshold and abnormal points with adjacent point velocity changes exceeding the acceleration threshold. The specific calculation formula is as follows: in, It represents the speed change rate between the ith operating point and the previous point, and eliminates those that satisfy | |> points, where It represents the acceleration threshold, which is used to eliminate abnormal points with sudden speed changes. When the density of valid points is lower than a specific density threshold, data compensation is performed on the virtual operating points generated based on the historical heat map in a proportional weighted manner. Assuming the screen area is S and the number of valid points is n, the density is: ,when Compensation is triggered when It represents the density threshold and stores the behavior data with spatiotemporal labels in a structured format, including timestamp, screen coordinates, and dwell time.

3. A shopping system based on WeChat applet according to claim 2, characterized in that: The number of times of re-adding after deletion is obtained by automatically associating the hot zone boundary with the product ID. When there is a hot zone covering the product card area and the stay time exceeds the first time threshold, the operation point is marked as a valid viewing behavior. For the comparison operation, the long press event of the multiple-choice button is listened to, and the timing starts from checking the first product and ends when the check is unchecked and the comparison is submitted. A comparison operation duration record is generated. All behavior data are sorted by timestamp and stored in the local cache. The number of views and the number of times of re-adding after deletion are accumulated with the product ID as the index.

4. A shopping system based on WeChat applet according to claim 2, characterized in that: The duration of the comparison operation is obtained by storing the complete time series record of the comparison operation event with the session ID as the index. When the density of valid points is insufficient, the virtual operation points generated based on the historical heat map will be assigned an associated ID according to the historical click-through rate of the product.

5. A shopping system based on WeChat applet according to claim 1, characterized in that: The three-dimensional orthogonal component decision vector generation module extracts features from the time and space dimensions. The time dimension calculates the operation frequency per unit time by counting the timestamp distribution density of valid operation points within the preset dynamic analysis window. The specific calculation formula is as follows: Among them, f represents the frequency of operations per unit time, which represents the degree of user activity. Indicates the number of valid operating points within the preset dynamic analysis window, Indicates the duration of the preset dynamic analysis window, which is used to characterize the activity level of user behavior, analyze the discreteness of operation point timestamps, and reflect the persistence of behavior by calculating the timestamp standard deviation; The spatial dimension calculates the overlapping area ratio of hot zones corresponding to different products. If the overlapping area exceeds a preset threshold, the product is determined to be strongly associated. The number of times users jump between different hot zones is counted to form a spatial jump frequency matrix for product operations. The features of the time dimension and the space dimension are input into the preset vector synthesis algorithm, and a decision vector containing three orthogonal components is output.

6. A shopping system based on WeChat applet according to claim 5, characterized in that: The first orthogonal component represents the purchase intention strength: the number of views in the window is accumulated by product ID to obtain the view density per unit time, the average length of stay of users in the product hot zone is calculated, and finally the concentration is measured by the distribution variance of the hot zone operation points. After standardizing the above three indicators, the purchase intention strength value is obtained.

7. A shopping system based on WeChat applet according to claim 5, characterized in that: The second orthogonal component characterizes price sensitivity: the cumulative number of times the same product is deleted and then added again during the memory period is counted, the time interval between each re-addition and deletion is calculated, and the maximum time interval is taken as a reference. The relationship between the number of times and the time interval is fitted through a logistic regression model to output the price sensitivity value.

8. A shopping system based on WeChat applet according to claim 5, characterized in that: The third orthogonal component characterizes the decision hesitation: by obtaining the duration of the comparison operation, and then calculating the standard deviation of the duration. The standard deviation reflects the time fluctuation of the user in different comparison sessions. If there are multiple comparison operations, further analyze the number of compared products and the number of hot zone jumps. The entropy weight method is used to calculate the weight of each indicator, and the decision hesitation value is obtained comprehensively. The three orthogonal components are synthesized into a three-dimensional orthogonal component decision vector in the orthogonal direction. ,in, The first, second, and third orthogonal components are represented and normalized respectively so that the vector modulus is in the interval [0, 1]. At the same time, a hot zone density attenuation factor is introduced. If the density of valid points in the current window is lower than the density threshold, the vector component needs to be multiplied by the attenuation coefficient to correct the impact of data compensation on the vector accuracy.

9. A shopping system based on WeChat applet according to claim 1, characterized in that: The settlement guidance strategy trigger module first builds a strategy template library and uses the cloud database to store three basic strategy templates, including Template A for issuing discount coupons, Template B for alerting for low inventory levels, and Template C for recommending similar products. The module then calculates the Euclidean distance between the decision vector of the three-dimensional orthogonal components and each template. The specific calculation formula is as follows: in, The Euclidean distance between the decision vector representing the three-dimensional orthogonal component and the k-th strategy template is used to measure the degree of matching. The smaller the value, the higher the matching degree. , , Indicates the preset component threshold of the kth strategy template, such as the price sensitivity threshold of template A and the purchase intention intensity threshold of template B. k represents the strategy template number. The template with the smallest distance is selected as the basic strategy, and the strategy parameters are adjusted according to the vector modulus. If the distances of multiple templates are all less than the threshold, they are weighted and combined according to the inverse of the distance, and the main strategy is displayed first.

10. A method for using a shopping system based on a mini-program is applied to a shopping system based on a WeChat mini-program as claimed in any one of claims 1 to 9, characterized in that: The following steps are involved: S101. Obtain shopping cart operation behavior sequence data through WeChat mini-program front-end event monitoring; S102, extracting time dimension and space dimension features, and outputting a decision vector of three-dimensional orthogonal components through a vector synthesis algorithm and orthogonal normalization; S103: Establish a policy template library, calculate the Euclidean distance between the decision vector and the template to match the policy, and adjust the strength parameter according to the vector modulus.

Images