Jewelry management system

Abstract

The invention discloses a jewelry management system, and particularly relates to the technical field of jewelry online shopping mall management, and the system comprises a personalized optimization module which analyzes page nodes and calculates the transition probability based on the historical browsing path data of a user, and dynamically adjusts the node sequence to optimize the user experience; the scene analysis module analyzes the complexity of a user browsing path and generates a behavior distribution coefficient and a path complexity coefficient; the fuzzy logic judgment module judges whether to activate the instantaneity optimization module or not by using fuzzy logic based on the analysis result; when the instantaneity optimization module is activated, page nodes are adjusted through user feedback, and the path display sequence is further optimized; the preloading module carries out preloading operation on page nodes in advance according to an optimization strategy, and the system response speed is increased; according to the method and the device, the next page node possibly accessed by the user can be intelligently predicted, and the corresponding page content is loaded in advance, so that the waiting time during page switching is shortened, and the loss of the user due to overlong waiting time is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of jewelry online mall management, and more specifically, to a jewelry management system. Background Art

[0002] In the current shopping environment, online jewelry stores have gradually become one of the main ways for users to buy jewelry. With the development of Internet technology and changes in consumer shopping behavior, online jewelry shopping has gradually surpassed the traditional offline physical store shopping method. Especially for high-end jewelry products, consumers expect to get a more intuitive experience and personalized recommendations during the browsing process to reduce browsing decision-making time and improve the shopping experience.

[0003] However, existing online jewelry stores still have certain limitations in terms of user experience and personalized recommendations. Most platforms provide static display pages, and the layout order and content display methods of the pages are usually fixed and will not be dynamically adjusted according to the user's browsing habits or preferences. This single display mode is difficult to fully meet the personalized needs of users, resulting in a relatively monotonous user experience and affecting purchase decisions. In addition, the delay problem of page loading speed often affects the user's browsing fluency, especially when the jewelry display content is more complex and contains high-quality pictures or videos. The user's waiting time is longer, which is easy to cause customer loss.

[0004] Therefore, there is an urgent need for a jewelry online mall system that can dynamically adjust the page display order according to user behavior and preferences, and at the same time improve the page response speed through intelligent preloading technology to optimize the user's browsing and shopping experience. Summary of the invention

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A jewelry management system, including a personalized optimization module, a scene analysis module, a fuzzy logic judgment module, an instant optimization module, and a preloading module;

[0007] The personalized optimization module analyzes the page nodes browsed by users based on their historical browsing path data, calculates the transition probability between each page node, and dynamically adjusts the order of nodes to optimize the user experience;

[0008] The scenario analysis module is used to analyze the scenario complexity of the user's browsing path, and generate the user's behavior distribution coefficient and path complexity coefficient based on the user's behavior information and scenario usage information;

[0009] The fuzzy logic judgment module uses fuzzy logic to judge whether to activate the immediacy optimization module based on the analysis results of the scenario analysis module;

[0010] When the immediacy optimization module is activated, it adjusts the page nodes based on the user's immediacy feedback behavior and further optimizes the path display order;

[0011] The preloading module is used to perform page node preloading operations according to the optimization strategy provided by the personalized optimization module or the instant optimization module.

[0012] When the personalized optimization module is working, each page node represents a state. Assume that there are N page nodes, namely S1, S2, ..., Sn, and the user's browsing history path is the transition process between one state and another.

[0013] In a preferred embodiment, calculating the transition probability between each page node refers to:

[0014] According to the user's historical browsing data, assuming that Pij represents the probability of transitioning from state Si to state Sj, the transition probability matrix P is:

[0015] i, j and n are all state numbers, i.e. page node indexes;

[0016] The formula for calculating the probability of transferring from one page node to another is:

[0017] CF si Indicates the number of departures from state Si, DD sj Indicates the number of transitions to state Sj.

[0018] In a preferred embodiment, dynamically adjusting the order of nodes refers to:

[0019] Set a steady-state distribution vector π, which represents the long-term probability distribution of each state when the system reaches a steady state. Vector π=(π1,π2,...,πn), where each element πi represents the probability that the system stays in state Si in the long run. In the steady-state distribution, π satisfies the following conditions:

[0020] πP=π;

[0021] n represents the total number of elements πi;

[0022] According to the steady-state distribution results, the nodes of the browsing path are sorted in descending order according to the size of πi, and the corresponding page nodes are displayed in the browsing path according to the results of the descending order.

[0023] In a preferred embodiment, the logic for obtaining the user behavior distribution coefficient is:

[0024] Get the probability distribution values ​​G(S1), G(S2), ..., G(Sn) corresponding to the user's historical behaviors on page nodes S1, S2, ..., Sn respectively, and then substitute them into the following formula:

[0025] wi is the purchase conversion rate corresponding to the preset page node Si, G(Si) represents the probability distribution value corresponding to the user's historical behavior on the page node Si, and FB represents the user behavior distribution coefficient.

[0026] In a preferred embodiment, the acquisition logic of the probability distribution value G(Si) is:

[0027] Obtain the historical behavior data of the user on the page node Si in the preset time window, extract the average time the user stays on the page node Si and record it as Ti, the average number of clicks of the user on the page node Si and record it as Ci, and the average scrolling distance of the user on the page node Si and record it as Ri;

[0028] Calculate the user's behavior value on page node Si:

[0029]

[0030] Xi represents the behavior value of the user on the page node Si, α, β, and γ are all preset non-zero influence coefficients, and the sum of all influence coefficients is one;

[0031] Calculate the probability distribution value G(Si):

[0032] In a preferred embodiment, the logic for obtaining the path complexity coefficient is:

[0033] Get the total path length from the start point to the end point of the user in the preset time window: di represents the distance from the starting point to the end point when the i-th page node is switched, m represents the total number of page node switches, and L represents the total path length;

[0034] Count the total number of all page nodes that the user passes through from the starting point to the end point, and when counting, count all repeated page nodes to obtain the total number of page nodes Y;

[0035] The formula for calculating the path complexity coefficient is:

[0036] f1, f2, and f3 are all preset non-zero proportional coefficients, T is the duration corresponding to the preset time window, and FZ is the path complexity coefficient.

[0037] In a preferred embodiment, the fuzzy logic judgment module uses fuzzy logic to judge whether to activate the immediacy optimization module based on the analysis result of the scenario analysis module, which means:

[0038] The user's behavior distribution coefficient and path complexity coefficient are taken as input variables, and the activation or not of the immediacy optimization module is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the degree of activation demand under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the result of whether the immediacy optimization module is activated.

[0039] In a preferred embodiment, when the immediacy optimization module is activated, the page nodes are adjusted according to the user's immediacy feedback behavior, and the path display sequence is further optimized by:

[0040] Using reinforcement learning algorithms, path optimization is performed based on the current usage status, i.e., the page node, the action, i.e., the next node to jump to, and the reward function, i.e., the user's interactive behavior.

[0041] Technical effects and advantages of the present invention:

[0042] The present invention analyzes the user's historical behavior data and instant feedback, dynamically adjusts the page display order, and enables the user to more quickly find the jewelry products they are interested in. This personalized display method greatly improves the user's browsing experience and makes the shopping process smoother and more efficient.

[0043] Through the preloading module, the present invention can intelligently predict the next page node that the user may visit, and load the corresponding page content in advance, reducing the waiting time when switching pages. This significantly improves the response speed of the system, improves the fluency of users when browsing high-quality pictures and videos, and avoids the loss of users due to long waiting time.

[0044] The present invention uses an immediacy optimization module to perform real-time analysis and path optimization based on the user's current browsing behavior, and can provide more personalized recommended content during the user's browsing process. This not only improves user participation, but also increases user stay time, and improves the display effect and purchase conversion rate of jewelry products.

[0045] The present invention uses a fuzzy logic judgment module, combined with the user behavior distribution coefficient and the path complexity coefficient, to intelligently determine whether to activate the immediacy optimization module. The system can flexibly adjust the optimization strategy according to the user's behavior, ensuring that resources can be reasonably allocated in scenarios of different complexity, thereby improving the overall performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to facilitate understanding by those skilled in the art, the present invention is further described below in conjunction with the accompanying drawings;

[0047] Figure 1 The figure is a schematic diagram of a jewelry management system in the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] Example 1

[0050] A jewelry management system, including a personalized optimization module, a scene analysis module, a fuzzy logic judgment module, an instant optimization module, and a preloading module;

[0051] The personalized optimization module analyzes the user's behavior at different page nodes based on the user's historical browsing path data and calculates the transition probability between these pages. Based on this data, the display order of page nodes is dynamically adjusted to optimize the user experience. The core function of this module is to provide a customized page display order based on the user's personalized behavior (such as browsing path, dwell time, etc.), so that the page content seen by the user is more in line with his or her personal interests and behavioral habits, thereby improving the user experience and participation.

[0052] The scenario analysis module analyzes the scenario complexity of the user's current browsing path. It generates the user's behavior distribution coefficient and path complexity coefficient based on the user's behavior information (such as dwell time, etc.) and scenario usage information (such as the number of page nodes, etc.). The scenario analysis module is used to identify whether the user's current browsing scenario is complex, such as whether the user browses multiple pages or how long the user stays. By generating the user's behavior distribution coefficient and path complexity coefficient, the system can determine the complexity of the current browsing path and provide basic data for subsequent optimization.

[0053] The fuzzy logic judgment module uses fuzzy logic to determine whether to activate the immediacy optimization module based on the analysis results provided by the scenario analysis module. If the overall complexity of the scenario is high, the module will start the immediacy optimization. The role of the fuzzy logic judgment module is to solve black-and-white decision-making problems. It can make fuzzy judgments between user behavior data and scenario complexity. If the scenario is complex, the system can use fuzzy logic to determine whether further optimization is needed to improve the user experience.

[0054] When activated by the fuzzy logic judgment module, the immediacy optimization module dynamically adjusts and optimizes the display order of page nodes through the user's immediate feedback behavior (such as clicks, slides, and dwell time). The immediacy optimization module is a module that the system responds quickly to the user's immediate behavior. When the system detects that the user's behavior is abnormal or the scene complexity is high, the module will respond immediately and optimize and adjust the page nodes according to the user's immediate feedback. This helps to improve the flexibility of the system and user satisfaction, so that users can still get a good experience in complex scenarios.

[0055] The preloading module preloads the page nodes that users may visit in advance according to the optimization strategy provided by the personalized optimization module or the immediacy optimization module to reduce waiting time. The preloading module is designed to improve the response speed and fluency of the system. By predicting the pages that users may visit, the system can load these pages into the cache in advance to avoid long loading times when users click on pages, thereby improving the immediacy and smoothness of the user experience. When the personalized optimization module is working, each page node represents a state. Assuming there are N page nodes, namely S1, S2, ..., Sn, the user's browsing history path is the transition process between one state and another.

[0056] Calculating the transition probability between each page node means:

[0057] According to the user's historical browsing data, assume that Pij represents the probability of transitioning from state Si to state Sj. This probability is based on the user's historical browsing data and reflects the user's behavioral tendency between different page nodes. The matrix P is the set of transition probabilities between all page nodes. The dimension of the matrix is ​​n×n, where n is the total number of page nodes. The transition probability matrix P is:

[0058] i, j and n are all state numbers, i.e., page node indexes. The rows represent the current page node Si, and the columns represent the target page node Sj;

[0059] The formula for calculating the probability of transferring from one page node to another is:

[0060] CF Si Indicates the number of departures from state Si, DD SjRepresents the number of transitions to state Sj. This formula calculates the probability of jumping from page node Si to page node Sj, reflecting the user's habit of jumping between different pages. The transition probability matrix P is used to comprehensively describe the user's jumping behavior between different pages. Each element Pij in the matrix represents the probability of transitioning from page Si to page Sj. The transition probability Pij represents the possibility of a user jumping from one page to another, which is calculated based on the number of jumps in historical data. This probability can be used to analyze the user's page preferences and behavior patterns, which is helpful for personalized recommendations and page optimization. The transition probability matrix and formula can be used to optimize the user's browsing path and dynamically adjust the display order of page nodes, so that the system can provide users with content that is more in line with their interests and improve the user experience. By calculating the transition probability, the system can better understand the user's behavior path between different pages and use this information to optimize page display and preloading strategies.

[0061] Dynamically adjusting the order of nodes means: setting a steady-state distribution vector π, which represents the long-term probability distribution of each state when the system reaches a steady state. Vector π = (π1, π2, ..., πn), where each element πi represents the probability that the system stays in state Si in the long run. In the steady-state distribution, π satisfies the following conditions:

[0062] πP=π; means that when the system reaches a steady state, the probability of the system changing from one state (page node) to another will not change. This formula shows that the steady-state distribution π is the eigenvector of the transition probability matrix P. This formula is used to calculate the stable distribution of each page node in the long-term operation of the system. Through this formula, the system can predict the behavior pattern of users in long-term visits.

[0063] n represents the total number of elements πi; this formula represents the normalization condition of the steady-state distribution vector, indicating that the sum of the probabilities of all states (page nodes) is 1. Ensure that the total probability of the system in all page nodes is 100%, that is, the user will definitely stay on a certain page node.

[0064] According to the steady-state distribution results, the nodes of the browsing path are sorted in descending order according to the size of πi, and the corresponding page nodes are displayed in the browsing path according to the results of the descending sorting. After the system has been running for a long time, the steady-state distribution vector π describes which page nodes the user is more likely to stay on, and the pages with higher probabilities will have higher priorities. According to the steady-state distribution results, the nodes in the browsing path are sorted in descending order according to the size of πi, and the nodes with higher probabilities are displayed first. The larger the steady-state distribution value πi, the higher the probability that the page node Si will be visited. The system will give priority to displaying it in the browsing path to optimize the user's browsing experience. Through the calculation of the steady-state distribution vector, the system can dynamically adjust the display order of the page nodes, making it easier for users to access the pages they are interested in or frequently visited. This optimization can improve the user's browsing efficiency and the personalized recommendation effect of the system.

[0065] The logic for obtaining the user behavior distribution coefficient is as follows: obtain the probability distribution values ​​G(S1), G(S2), ..., G(Sn) corresponding to the user's historical behavior at page nodes S1, S2, ..., Sn respectively, and then substitute them into the following formula:

[0066] wi is the purchase conversion rate corresponding to the preset page node Si, G(Si) represents the probability distribution value corresponding to the user's historical behavior on the page node Si, and FB represents the user behavior distribution coefficient. It represents the probability distribution value corresponding to the user's historical behavior on the page node Si. These values ​​are based on the user's historical browsing data, such as click-through rate, dwell time, etc., and represent the user's behavioral preferences on different page nodes. By obtaining the user's behavior distribution on each page node, the system can understand the user's historical behavior characteristics on each page node. wi is the preset weight of the page node Si, which represents the conversion efficiency of the page node and the user's behavior (for example, the purchase conversion rate or the conversion rate of other target behaviors). The larger the weight, the higher the importance of the page node in the target behavior (such as purchase). In actual use, it can also be changed to the conversion rate of other target behaviors. The weight wi can be used to reflect the importance of the page node, so that important page nodes have a greater impact on the user behavior distribution coefficient FB. log e G(Si) represents the logarithm of the user behavior distribution G(Si). The more uncertain the behavior distribution, the higher the complexity. FB is the overall behavior distribution coefficient calculated based on the historical behavior data of users on different page nodes, which represents the uncertainty of user behavior between various page nodes. The larger the FB, the higher the uncertainty of user behavior on each page node, the more complex the user's browsing path, or the user has a relatively uniform behavior distribution on multiple page nodes (that is, the user behavior is scattered among multiple nodes). This situation may mean that the user does not have a clear goal when browsing multiple pages, or that each page has a similar appeal to the user.

[0067] The logic for obtaining the probability distribution value G(Si) is:

[0068] Obtain the historical behavior data of users on page node Si in a preset time window, extract the average time users stay on page node Si and record it as Ti, the average number of clicks of users on page node Si and record it as Ci, and the average scrolling distance of users on page node Si and record it as Ri; these data are used to quantify the user's behavior on the page to reflect the user's attention to the page.

[0069] Calculate the user's behavior value on page node Si:

[0070]

[0071] Xi represents the user's behavior value on the page node Si. α, β, and γ are all preset non-zero influence coefficients, and the sum of all influence coefficients is one. The data is smoothed by taking the natural logarithm of the residence time to avoid the influence of extremely long time. The square root of the number of clicks is taken to balance the situation where the number of clicks between different pages is large. The square of the scrolling distance is to amplify the impact of scrolling behavior on user behavior. This formula is used to comprehensively measure the user's behavior performance on the page and reflect the user's attention to the page node Si.

[0072] Calculate the probability distribution value G(Si): G(Si) represents the proportion of user behavior on page Si relative to other pages. This relative value reflects the user's preference for different pages, and through normalization, the sum of the distribution values ​​of all page nodes is ensured to be 1. The larger G(Si) is, the more active the user's behavior is on page Si.

[0073] The logic for obtaining the path complexity coefficient is: in the preset time window, obtain the total path length that the user has traveled from the starting point to the end point: di represents the distance when the page node is switched for the i-th time from the starting point to the end point, m represents the total number of page node switches, and L represents the total path length; the path length L reflects the overall length of the user's browsing path within a period of time. The more times the user switches between pages, the more complex the path.

[0074] The total number of all page nodes that the user passes through from the starting point to the end point is counted, and when counting, all repeated page nodes are counted to obtain the total number of page nodes Y; the total number of page nodes Y reflects the types of pages visited in the user's browsing path, especially considering the repeatedly visited pages, which can further reflect the user's interests and behavior complexity.

[0075] The formula for calculating the path complexity coefficient is:

[0076] f1, f2, and f3 are all preset non-zero proportional coefficients, which respectively adjust the path length, the number of page nodes, and the impact of time on the path complexity coefficient. The sum of the three as the denominator also meets the requirement of not being zero. T is the duration corresponding to the preset time window, and FZ is the path complexity coefficient.

[0077] The complexity of user browsing behavior is comprehensively measured by considering the number of times users switch between pages (path length), the number of pages visited (including pages visited repeatedly), and the time users spend. The more complex the path, the more frequent the jump, the more pages visited, or the longer the time spent by the user during browsing. The time factor T is used in this formula to reflect the time window of user browsing. If a user visits more pages in a shorter period of time, it may indicate that the user's path complexity is high, and the system can be optimized based on this indicator. By adjusting f1, f2, and f3, the system can adapt to different scenarios. For example, in some cases, the path length L may be the most important factor in evaluating complexity, while in other cases, the number of pages Y or the time T may be more important.

[0078] The larger the FZ, the more complex the user's browsing path is, which may include a longer browsing path or more page nodes. This means that the user's behavior pattern is more complex and further analysis of user behavior may be needed to provide more accurate recommendations or optimization solutions.

[0079] The fuzzy logic judgment module uses fuzzy logic to determine whether to activate the immediacy optimization module based on the analysis results of the scenario analysis module:

[0080] The user's behavior distribution coefficient and path complexity coefficient are taken as input variables, and the activation or not of the immediacy optimization module is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the degree of activation demand under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the result of whether the immediacy optimization module is activated.

[0081] The behavior distribution coefficient reflects the complexity and distribution of user behaviors on each page node, indicating the differences in user behaviors on different pages. The path complexity coefficient measures the complexity of the user's browsing path, including the path length, the number of page nodes, and the time window factor, reflecting the complexity of the user's browsing behavior. These two input variables will be used to determine whether immediacy optimization is needed. The input variables are fuzzified, that is, the specific values ​​of the user's behavior distribution coefficient and the path complexity coefficient are converted into fuzzy sets. The purpose of this step is to convert precise values ​​into fuzzy semantics such as "low", "medium", and "high".

[0082] Fuzzification of the behavior distribution coefficient: For example, the values ​​of the behavior distribution coefficient are divided into three fuzzy sets: "low", "medium", and "high". The greater the difference in user behavior on different pages, the higher the value of the behavior distribution coefficient. Fuzzification of the path complexity coefficient: For example, the values ​​of the path complexity coefficient are divided into three fuzzy sets: "simple", "moderate", and "complex". The more times the user jumps to the page during browsing and the longer the time, the higher the value of the path complexity coefficient. Fuzzification of the output variable, that is, the activation of the immediacy optimization module is divided into different fuzzy sets.

[0083] Output variable (activated or not): can be divided into three fuzzy sets of "low demand", "medium demand" and "high demand", which represent the degree of demand for activating the immediacy optimization module. The higher the activation demand, the more likely the system is to start immediacy optimization.

[0084] Formulate fuzzy rules: Fuzzy rules are response settings for different combinations of input variables. These rules are used to determine whether to activate the immediacy optimization module. The input of each rule is a different combination of behavior distribution coefficient and path complexity coefficient, and the output is the degree of activation demand.

[0085] Fuzzy rule example: If the behavior distribution coefficient is "high" and the path complexity coefficient is "complex", the activation demand is "high demand". If the behavior distribution coefficient is "medium" and the path complexity coefficient is "moderate", the activation demand is "medium demand". If the behavior distribution coefficient is "low" and the path complexity coefficient is "simple", the activation demand is "low demand". These rules reflect the relationship between the complexity of user behavior and the complexity of browsing paths, helping the system make timely optimization judgments.

[0086] Through fuzzy rules, the fuzzified input variables are inferred. According to different combinations of behavior distribution coefficients and path complexity coefficients, fuzzy reasoning will calculate the degree of activation demand for the immediacy optimization module. Fuzzy logic combines the fuzzy values ​​of input variables and calculates according to the set fuzzy rules to obtain the possibility of activating the immediacy optimization module. Through the calculation of fuzzy reasoning, the system will ultimately determine whether the immediacy optimization module needs to be activated. If the reasoning result is "high demand", the immediacy optimization module is activated; if it is "medium demand" or "low demand", it is not activated.

[0087] When the immediacy optimization module is activated, the page nodes are adjusted according to the user's immediate feedback behavior, and the path display sequence is further optimized. The operations are:

[0088] Using reinforcement learning algorithms, path optimization is performed based on the current usage status, i.e., the page node, the action, i.e., the next node to jump to, and the reward function, i.e., the user's interactive behavior.

[0089] In the immediacy optimization module, the application of reinforcement learning algorithm optimizes the path display order of page nodes through the user's immediate feedback behavior, ensuring that the user's browsing experience is more efficient and personalized.

[0090] In reinforcement learning, the system gradually optimizes decisions through interaction with the environment (i.e., the user's browsing behavior). The core idea of ​​reinforcement learning is to gradually learn the optimal behavior strategy based on the "state-action-reward" loop feedback mechanism. The following is an explanation of its main principles:

[0091] State: In this scenario, the state refers to the page node that the user is currently browsing. For example, if the user stays on a specific page node, this page node represents the current state of the system.

[0092] Action: An action refers to an operation that a user may take in the current state. In this scenario, an action refers to the next page node that a user may jump to from the current page. The system can predict the user's potential jump behavior through a reinforcement learning algorithm and select the optimal path.

[0093] Rewards: Rewards are feedback obtained by the system from user interaction behaviors, which are used to guide path optimization. In this scenario, user behaviors (such as clicks, dwell time, and scrolling distance) will be used as the basis for rewards. Positive behavioral feedback (such as users staying for a long time or having a high click rate) will bring positive rewards, while negative behavioral feedback (such as users leaving quickly or not interacting) may bring negative rewards.

[0094] Value function and strategy: The goal of reinforcement learning is to learn an optimal strategy, that is, to select the best action in each state to maximize the long-term accumulated rewards. Through continuous trial and error, the system can gradually optimize the path display order, so that users can get the best page experience in complex browsing scenarios.

[0095] Q-value update: Determine the value of taking an action from a certain state by learning a state-action value (i.e., Q-value). In this system, Q-value can be understood as the expected benefit of jumping from the current page node to the next page node. The Q-value update formula is:

[0096] s is the current page node, a is the next node to jump to from the current page; r is the reward obtained from the user's immediate feedback (such as clicks, dwell time); α is the learning rate, which controls the speed of updates; γ is the discount factor, which indicates the impact of future rewards on current decisions. It represents the maximum Q value, i.e. the maximum future benefit, among all possible actions a′ at the next page node s′.

[0097] Each time a user browses, the system records the path the user chooses from one page node to another and updates the Q value based on the user's immediate feedback. As the number of interactions increases, the system gradually learns to select the optimal page display order under different conditions and optimizes the user's browsing path. Ultimately, users can find content they are interested in more quickly on pages that have undergone reinforcement learning, thereby improving the system's response speed and user experience.

[0098] The preloading module is used to perform page node preloading operations according to the optimization strategy provided by the personalized optimization module or the instant optimization module.

[0099] Strategies provided by the personalized optimization module: Prioritize page loading based on user history data: The personalized optimization module analyzes the user's historical browsing path data, calculates the transfer probability of each page node, and dynamically adjusts the order of the nodes. This optimization strategy determines which pages the user is more likely to visit. The preloading module will load the page nodes that the user may jump to in advance based on this optimization result.

[0100] Strategies provided by the immediacy optimization module: Predictive loading based on user's immediate behavior: The immediacy optimization module uses a reinforcement learning algorithm to predict the next page a user may visit based on the user's current status and immediate feedback. The preloading module will load the pages that the user may click in advance based on this prediction strategy to reduce the jump time. When the user's behavior changes, the system will readjust the path display order based on the new feedback, and the preloading module will dynamically adjust the page nodes to be loaded to ensure consistency with the user's current behavior.

[0101] The preloading module can load the page that the user is about to visit in advance according to the optimization strategy, thereby reducing the waiting time of the user when jumping to the page and improving the system response speed. This is especially important in complex pages or pages with more resources, and can significantly improve the user experience.

[0102] By loading likely pages in advance, users will feel smoother during browsing and will not be interrupted by long page loading times. This seamless browsing experience helps increase user retention time in the system and reduces churn caused by page loading delays.

[0103] The system does not preload all pages, but intelligently selects the pages that users are most likely to visit for preloading through personalized and immediacy optimization strategies. This targeted preloading not only improves the user experience, but also reduces unnecessary system resource consumption.

[0104] Whether it is a personalized optimization strategy or an instant optimization strategy, the preloading module can dynamically adjust the loaded page content to ensure that users can get fast page response in different scenarios. This flexibility ensures that the system can respond to changes in user needs in a timely manner.

[0105] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0106] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0107] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0108] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0109] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A jewelry management system, characterized in that: Including personalized optimization module, scenario analysis module, fuzzy logic judgment module, instant optimization module, and preloading module; The personalized optimization module analyzes the page nodes browsed by users based on their historical browsing path data, calculates the transition probability between each page node, and dynamically adjusts the order of nodes to optimize the user experience; The scenario analysis module is used to analyze the scenario complexity of the user's browsing path, and generate the user's behavior distribution coefficient and path complexity coefficient based on the user's behavior information and scenario usage information; The fuzzy logic judgment module uses fuzzy logic to judge whether to activate the immediacy optimization module based on the analysis results of the scenario analysis module; When the immediacy optimization module is activated, it adjusts the page nodes based on the user's immediacy feedback behavior and further optimizes the path display order; The preloading module is used to perform page node preloading operations according to the optimization strategy provided by the personalized optimization module or the instant optimization module.

2. A jewelry management system according to claim 1, characterized in that: When the personalized optimization module is working, each page node represents a state. Assume that there are N page nodes, namely S1, S2, ..., Sn, and the user's browsing history path is the transition process between one state and another.

3. A jewelry management system according to claim 2, characterized in that: Calculating the transition probability between each page node means: According to the user's historical browsing data, assuming that Pij represents the probability of transitioning from state Si to state Sj, the transition probability matrix P is: i, j and n are all state numbers, i.e. page node indexes; The formula for calculating the probability of transferring from one page node to another is: CF Si Indicates the number of departures from state Si, DD Sj Indicates the number of transitions to state Sj.

4. A jewelry management system according to claim 3, characterized in that: Dynamically adjusting the order of nodes means: Set a steady-state distribution vector π, which represents the long-term probability distribution of each state when the system reaches a steady state. Vector π=(π1,π2,...,πn), where each element πi represents the probability that the system stays in state Si in the long run. In the steady-state distribution, π satisfies the following conditions: πP＝π； n represents the total number of elements πi; According to the steady-state distribution results, the nodes of the browsing path are sorted in descending order according to the size of πi, and the corresponding page nodes are displayed in the browsing path according to the results of the descending order.

5. A jewelry management system according to claim 4, characterized in that: The logic for obtaining the user behavior distribution coefficient is: Get the probability distribution values ​​G(S1), G(S2), ..., G(Sn) corresponding to the user's historical behaviors on page nodes S1, S2, ..., Sn respectively, and then substitute them into the following formula: wi is the purchase conversion rate corresponding to the preset page node Si, G(Si) represents the probability distribution value corresponding to the user's historical behavior on the page node Si, and FB represents the user behavior distribution coefficient.

6. A jewelry management system according to claim 5, characterized in that: The logic for obtaining the probability distribution value G(Si) is: Obtain the historical behavior data of the user on the page node Si in the preset time window, extract the average stay time of the user on the page node Si and record it as Ti, the average number of clicks of the user on the page node Si and record it as Ci, and the average scrolling distance of the user on the page node Si and record it as Ri; Calculate the user's behavior value on page node Si: Xi represents the behavior value of the user on the page node Si, α, β, and γ are all preset non-zero influence coefficients, and the sum of all influence coefficients is one; Calculate the probability distribution value G(Si):

7. A jewelry management system according to claim 6, characterized in that: The logic for obtaining the path complexity coefficient is: Get the total path length from the start point to the end point of the user in the preset time window: di represents the distance from the starting point to the end point when the i-th page node is switched, m represents the total number of page node switches, and L represents the total path length; Count the total number of all page nodes that the user passes through from the starting point to the end point, and when counting, count all repeated page nodes to obtain the total number of page nodes Y; The formula for calculating the path complexity coefficient is: f1, f2, and f3 are all preset non-zero proportional coefficients, T is the duration corresponding to the preset time window, and FZ is the path complexity coefficient.

8. A jewelry management system according to claim 7, characterized in that: The fuzzy logic judgment module uses fuzzy logic to determine whether to activate the immediacy optimization module based on the analysis results of the scenario analysis module: The user's behavior distribution coefficient and path complexity coefficient are taken as input variables, and the activation or not of the immediacy optimization module is taken as the output variable. The input variables are fuzzified and the values ​​of the input variables are converted into fuzzy sets. The output variables are fuzzified and the output variables are converted into fuzzy sets. Fuzzy rules are formulated to describe the degree of activation demand under different data type combinations. The fuzzified input variables are inferred through fuzzy rules to obtain the result of whether the immediacy optimization module is activated.

9. A jewelry management system according to claim 8, characterized in that: When the immediacy optimization module is activated, the page nodes are adjusted according to the user's immediate feedback behavior, and the path display sequence is further optimized. The operations are: Using reinforcement learning algorithms, path optimization is performed based on the current usage status, i.e., the page node, the action, i.e., the next node to jump to, and the reward function, i.e., the user's interactive behavior.

Images