Intelligent visitor information management system based on big data analysis

Abstract

The invention relates to the technical field of visitor information intelligent management, and particularly discloses a visitor information intelligent management system based on big data analysis, which comprises a visitor portrait generation module, a platform bearing capacity analysis module and a platform bearing early warning module, collecting multi-dimensional visitor information flow from a platform background database in real time; performing multi-modal fusion processing on the multi-dimensional visitor information flow, and constructing a dynamic portrait matrix of each visitor; analyzing the visitor dynamic portrait matrix based on a deep space-time convolutional network, and extracting resource demand characteristics of each visitor; generating a threat assessment index of the platform bearing capacity; and carrying out early warning management on the bearing capacity of the platform based on the threat evaluation index of the bearing capacity of the platform. And the understanding depth of the platform on visitor behaviors is greatly improved. The platform can integrate access behaviors of visitors, network protocol parameters, session tracks and other multi-source information, and comprehensively and stereoscopically depicts behavior characteristics and resource demands of each visitor, thereby realizing accurate identification and dynamic tracking of visitor demands.

Description

Technical Field

[0001] The present invention belongs to the technical field of visitor information intelligent management, and relates to a visitor information intelligent management system based on big data analysis. Background Art

[0002] With the rapid growth of internet visitors, platforms may suddenly face a surge in visitor traffic. If platforms are unable to effectively handle this surge, they can experience system overload, response delays, or even crashes, leading to severe visitor experience issues and business losses. Therefore, platform crash warnings based on visitor information are crucial. First, by real-time monitoring and analysis of visitor behavior data, access patterns, and visit volume, abnormal growth trends in visitor volume can be detected, predicting impending platform pressure peaks and providing a valuable window of opportunity to implement appropriate countermeasures. Second, platform crash warnings can help operators rationally plan resources, optimize system configurations, and adjust service strategies in a timely manner, effectively addressing the challenges posed by the surge in visitors and ensuring stable platform operation and service quality. This early warning mechanism not only prevents the negative impact of platform crashes but also improves the visitor experience, enhancing the platform's competitiveness and sustainable development capabilities.

[0003] However, current technologies still have some flaws and drawbacks in using visitor information to provide platform crash warnings when faced with a surge in visitors or an impending crash. First, some existing technologies fail to collect and analyze visitor information in a timely and comprehensive manner, resulting in insufficient accuracy and real-time nature of warnings and an inability to detect potential crash risks. Furthermore, existing technologies have limitations in integrating and analyzing visitor information, failing to fully explore the correlations and potential value between different data sources, impacting the accuracy and effectiveness of warning decisions. Summary of the Invention

[0004] In view of the above problems existing in the prior art, the present invention provides a visitor information intelligent management system based on big data analysis to solve the above technical problems.

[0005] In order to achieve the above-mentioned and other purposes, the technical solutions adopted by the present invention are as follows:

[0006] The present invention provides an intelligent visitor information management system based on big data analysis, which includes a visitor portrait generation module, a platform carrying capacity analysis module, and a platform carrying capacity warning module; wherein each module is connected by wired and / or wireless connection to realize data transmission between each module;

[0007] Visitor portrait generation module: collects multi-dimensional visitor information flows from the platform backend database in real time; performs multi-modal fusion processing on the multi-dimensional visitor information flows to construct a dynamic portrait matrix for each visitor;

[0008] Platform carrying capacity analysis module: Analyzes the visitor dynamic profile matrix based on a deep spatiotemporal convolutional network to extract the resource demand characteristics of each visitor; and then generates a threat assessment index for the platform's carrying capacity;

[0009] Platform carrying capacity warning module: Based on the threat assessment index of the platform carrying capacity, it provides early warning management of the platform's carrying capacity.

[0010] For example, multi-modal fusion processing is performed on the multi-dimensional visitor information flow to construct a dynamic portrait matrix of each visitor. The specific construction process is as follows:

[0011] The multi-dimensional visitor information flow includes the visit behavior data, network protocol data and session trajectory data of each visitor;

[0012] Performing multi-source spatiotemporal benchmark alignment processing on the access behavior data, network protocol data, and session trajectory data to eliminate timestamp drift of heterogeneous data streams and generate multimodal fusion base data for each visitor;

[0013] Performing three-dimensional feature decoupling on the multimodal fusion base data to generate visit behavior attribute data, network protocol data, and session trajectory sequence data of each visitor;

[0014] Based on the access behavior attribute data, cross-modal correlation analysis is performed on the session trajectory sequence data and the network protocol data, dynamic coupling weights are calculated through a bidirectional attention mechanism, and a behavior-protocol-trajectory correlation matrix of each visitor is generated;

[0015] The behavior-protocol-trajectory association matrix is ​​dynamically fused in time and space to construct a dynamic portrait matrix of each visitor.

[0016] Exemplarily, the three-dimensional feature decoupling is performed on the multimodal fusion base data, and the specific logic is as follows:

[0017] Establish an operation type-frequency association matrix, count the operation type distribution of each visitor according to the preset time window, normalize the original frequency values ​​by standard deviation, assign a weight coefficient of 2 times to high-frequency abnormal operation types, and generate a weighted operation entropy value. Simultaneously, use a sliding window mechanism to construct an operation time series histogram, calculate the coefficient of variation of the proportion of the main operation types within the window, and finally linearly combine the weighted entropy value and the coefficient of variation to output a composite attribute indicator representing the concentration of each visitor's behavior. This is used as the visit behavior attribute data of each visitor.

[0018] A pattern recognition algorithm is used to analyze the sequence of pages continuously visited by each visitor, and the set of frequent jump paths that meet the minimum support threshold is counted. A probability distribution is fitted for the samples of the dwell time of each page type, and the shape parameter α and scale parameter β are extracted. A state transition matrix between pages is constructed, and the number of jumps from the current interface to the possible next interface {P1, P2, ... Pn} by each visitor is counted. The probability value of each jump direction is calculated to form a discrete probability distribution. Based on the discrete characteristics of the probability distribution, the Shannon entropy formula is applied to calculate its entropy value, which is used to quantify the degree of certainty of each visitor's navigation path, and then the session trajectory sequence data of each visitor is determined.

[0019] Exemplarily, the dynamic spatiotemporal fusion processing is performed on the behavior-protocol-trajectory association matrix, and the specific processing process is as follows:

[0020] Preprocessing the correlation matrix within a preset time sliding window to calculate the energy spectrum density of each visitor;

[0021] Based on the energy spectrum density, entropy weights are dynamically aggregated to generate the spatiotemporal synchronization feature tensor of each visitor;

[0022] The spatiotemporal synchronization feature tensor is input into the feature projection space to construct a dynamic portrait matrix of each visitor.

[0023] For example, the resource demand characteristics of each visitor are extracted, and the specific extraction logic is as follows:

[0024] Performing a three-dimensional time window sliding dimensionality reduction on the spatiotemporal synchronization feature tensor based on a deep spatiotemporal convolutional network to obtain spatiotemporal feature dimensionality reduction data of each visitor;

[0025] Based on the spatiotemporal feature dimensionality reduction data, the topological clustering of mobile trajectories is identified to obtain clustered behavior trajectory data of each visitor;

[0026] Dynamically segment the behavior sequence of cluster behavior trajectory data to obtain the spatiotemporal distribution matrix of each visitor's behavior interval;

[0027] Based on the spatiotemporal distribution matrix of behavior intervals, a normal behavior baseline pattern is constructed to obtain the probability data of abnormal behavior deviation for each visitor;

[0028] A hidden Markov model of resource demand is established based on the abnormal behavior deviation probability data to extract resource demand characteristics of each visitor.

[0029] For example, the logic for obtaining cluster behavior trajectory data of each visitor is as follows:

[0030] The spectral clustering algorithm is used to detect the peak value of trajectory density of each visitor's spatiotemporal feature dimensionality reduction data and generate cluster center points;

[0031] The morphological similarity between each trajectory segment and the cluster center is calculated based on the dynamic time warping algorithm to obtain the trajectory membership matrix of each visitor;

[0032] Performing a spatiotemporal coupling analysis on the trajectory membership matrix to generate a group behavior coverage heat map of each visitor;

[0033] The behavioral clustering boundary lines are delineated according to the energy distribution density in the heat map to form clustering data of each visitor's cluster behavior trajectory.

[0034] Exemplarily, generating a threat assessment index of platform carrying capacity includes:

[0035] The access path graph is modeled based on the resource demand characteristic parameters of each visitor, and the path node influence factor is calculated using the directed graph betweenness centrality algorithm to generate the access path correlation quantification value of each visitor;

[0036] Based on the resource demand characteristic parameters of each visitor, service call sequence pattern mining is carried out, service combination dependency is calculated, and the service dependency quantification value of each visitor is generated;

[0037] Perform concurrent operation simulation on the clustered behavior trajectory data of each visitor, calculate the critical resource competition probability of each visitor, and generate the concurrent impact risk value of each visitor;

[0038] Based on the entropy weight method, dynamic weight allocation is performed on the visitor's access path correlation quantification value, service dependency quantification value and concurrent impact risk value to obtain a three-dimensional indicator weight matrix;

[0039] The three-dimensional indicator weight matrix is ​​normalized through the indicator fusion function, and the Euclidean distance value of the threat assessment index of each visitor is calculated, and then the threat assessment index of the platform carrying capacity is comprehensively generated.

[0040] For example, early warning management of the platform's carrying capacity is performed as follows:

[0041] Compare the threat assessment index of the platform's carrying capacity with the preset threat assessment threshold range [S1, S2];

[0042] If the threat assessment index of the platform's carrying capacity is less than S1, the platform is judged to be in a low-risk state and the conventional monitoring strategy is implemented;

[0043] If the threat assessment index of the platform's carrying capacity is greater than S2, the platform is judged to be in a high-risk state, triggering an emergency warning response and immediately activating the resource elastic expansion mechanism;

[0044] If the threat assessment index of the platform's carrying capacity is within the threat assessment threshold range, the platform is judged to be in a medium-risk state, and the dynamic load balancing strategy is activated to optimize resource allocation priority.

[0045] As described above, the visitor information intelligent management system based on big data analysis provided by the present invention has at least the following beneficial effects:

[0046] The visitor information intelligent management system based on big data analysis provided by the present invention collects multi-dimensional visitor information streams from the platform's backend database in real time, performs multimodal fusion processing on this information, constructs a dynamic portrait matrix, and uses a deep spatiotemporal convolutional network to analyze the characteristics of visitor resource demand, thereby generating a threat assessment index for the platform's carrying capacity and performing early warning management based on this. It has significant technical advantages and necessity. It greatly enhances the platform's understanding of visitor behavior. Through multimodal fusion, the platform can integrate multi-source information such as visitor access behavior, network protocol parameters, and session trajectories to comprehensively and three-dimensionally characterize each visitor's behavioral characteristics and resource requirements, thereby achieving accurate identification and dynamic tracking of visitor needs. This fusion of multi-dimensional information and dynamic portrait construction breaks the limitations of traditional single indicator monitoring, improves the ability to perceive complex visitor behavior, and provides a solid foundation for subsequent intelligent analysis. It not only greatly enhances the platform's understanding of visitor behavior and its ability to predict resource demand, but also significantly improves the platform's intelligent management level and emergency response capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 Schematic diagram of the connection of various modules of the system of the present invention. DETAILED DESCRIPTION

[0049] The above contents described below in conjunction with the implementation of the present invention are merely examples and explanations of the concept of the present invention. Those skilled in the art may make various modifications or additions to the described specific embodiments or replace them in a similar manner. As long as they do not deviate from the concept of the invention or exceed the scope defined by the claims, they shall fall within the scope of protection of the present invention.

[0050] Example 1

[0051] See also Figure 1 As shown, a visitor information intelligent management system based on big data analysis includes a visitor portrait generation module, a platform carrying capacity analysis module, and a platform carrying capacity warning module; wherein each module is connected by wired and / or wireless connection to realize data transmission between each module;

[0052] Visitor portrait generation module: collects multi-dimensional visitor information flows from the platform backend database in real time; performs multi-modal fusion processing on the multi-dimensional visitor information flows to construct a dynamic portrait matrix for each visitor;

[0053] Perform multimodal fusion processing on the multi-dimensional visitor information flow to construct a dynamic portrait matrix for each visitor. The specific construction process is as follows:

[0054] The multi-dimensional visitor information flow includes the visit behavior data, network protocol data and session trajectory data of each visitor;

[0055] It should be added that by analyzing the platform server logs, information such as each visitor's access time, visited pages, dwell time, and access source can be obtained. Network traffic data can be captured through network packet capture tools, including but not limited to IP addresses, port numbers, protocol types, etc. Use visitor behavior tracking tools to record the behavioral path and page browsing sequence of each visitor on the platform. The access behavior data includes but is not limited to access time, visited page URL, dwell time, access source (Referrer), visitor device information, number of page views, etc. Network protocol data includes but is not limited to IP address, port number, protocol type, packet size, transmission rate, transmission direction, packet flag, etc. Session track data includes but is not limited to session ID, access time, page path, page jump, dwell time, interactive actions (click, scroll), conversion behavior (registration, purchase), etc.

[0056] Performing multi-source spatiotemporal benchmark alignment processing on the access behavior data, network protocol data, and session trajectory data to eliminate timestamp drift of heterogeneous data streams and generate multimodal fusion base data for each visitor;

[0057] Performing three-dimensional feature decoupling on the multimodal fusion base data to generate visit behavior attribute data, network protocol data, and session trajectory sequence data of each visitor;

[0058] Based on the access behavior attribute data, cross-modal correlation analysis is performed on the session trajectory sequence data and the network protocol data, dynamic coupling weights are calculated through a bidirectional attention mechanism, and a behavior-protocol-trajectory correlation matrix of each visitor is generated;

[0059] The behavior-protocol-trajectory association matrix is ​​dynamically fused in time and space to construct a dynamic portrait matrix of each visitor.

[0060] This embodiment of the present invention deploys a distributed clock synchronization subunit in the data processing server. It first performs a unified benchmark calibration on the original timestamps of each data source. Using a sliding window phase alignment algorithm, it performs cross-device time correction on the request event timestamps of access behavior data, packet arrival timestamps of network protocol data, and interface jump timestamps of session trajectory data in heterogeneous data streams.

[0061] A three-dimensional feature cube is constructed in the platform association analysis server, and the access behavior attributes, protocol sequence features, and session trajectory features are mapped to three orthogonal dimensions of height, width, and depth respectively. The association weight of the access behavior to the protocol features is first calculated along the height dimension, and the dot product attention scoring function is used to measure the matching degree between the specific operation type and the protocol feature. The spatiotemporal association between the session trajectory and the protocol features is established along the depth dimension. By setting a model to capture the dynamic coupling relationship between the page dwell time and the TCP window size, a time-attenuated weighted association probability matrix is ​​generated. Finally, the three-dimensional association weights are fused, and a tensor shrinking operation is performed on the feature cube to output an N×N behavior-protocol-trajectory association matrix (N is the total number of visitors), where each element represents the comprehensive association strength between two visitors under the three modes.

[0062] The three-dimensional feature decoupling of the multimodal fusion base data is performed, and the specific logic is as follows:

[0063] Establish an operation type-frequency association matrix, count the operation type distribution of each visitor according to the preset time window, normalize the original frequency values ​​by standard deviation, assign a weight coefficient of 2 times to high-frequency abnormal operation types, and generate a weighted operation entropy value. Simultaneously, use a sliding window mechanism to construct an operation time series histogram, calculate the coefficient of variation of the proportion of the main operation types within the window, and finally linearly combine the weighted entropy value and the coefficient of variation to output a composite attribute indicator representing the concentration of each visitor's behavior. This is used as the visit behavior attribute data of each visitor.

[0064] A pattern recognition algorithm is used to analyze the sequence of pages visited by each visitor, counting frequent jump paths that meet a minimum support threshold (e.g., appearing more than 50 times per day). For example, a typical navigation link, "Home → Product Category → Product Details → Shopping Cart," is identified. A probability distribution is fitted for the dwell time samples of each page type, using a Weibull distribution model to extract the shape parameter α (representing the central tendency of the time distribution) and the scale parameter β (representing the average dwell time). For example, for the product details page, α = 1.5 and β = 58 seconds are fitted. A state transition matrix between pages is constructed, counting the number of jumps from the current page to the possible next pages {P1, P2, ...Pn}, and calculating the probability of each jump direction to form a discrete probability distribution. Based on the discrete nature of the probability distribution, the Shannon entropy formula is used to calculate its entropy value, which is used to quantify the degree of certainty of each visitor's navigation path, thereby determining the sequence data of each visitor's session trajectory.

[0065] The behavior-protocol-trajectory association matrix is ​​subjected to dynamic spatiotemporal fusion processing. The specific processing process is as follows:

[0066] Preprocessing the correlation matrix within a preset time sliding window to calculate the energy spectrum density of each visitor;

[0067] Based on the energy spectrum density, entropy weights are dynamically aggregated to generate the spatiotemporal synchronization feature tensor of each visitor;

[0068] The spatiotemporal synchronization feature tensor is input into the feature projection space to construct a dynamic portrait matrix of each visitor.

[0069] In the embodiment of the present invention, the time sliding window parameters are pre-set. For each visitor's behavior-protocol-trajectory association matrix within the window period, a sliding accumulation process of the time series features is first performed: for the parameters in the association matrix within the current window, normalization preprocessing is first performed to make their values ​​in the [0,1] interval, and then they are converted into frequency domain feature representations through fast Fourier transform. The energy spectrum density calculation adopts a frequency band integration strategy, dividing the transformed spectrum into three characteristic frequency bands: low frequency, medium frequency, and high frequency. The sum of the absolute energy values ​​of each frequency band is calculated respectively, and the energy spectrum density of each visitor is converted by the frequency band energy ratio.

[0070] The ratio of the standard deviation to the mean of the low-frequency, medium-frequency, and high-frequency energy values ​​of each visitor during the sliding window period is calculated as the coefficient of variation, and the weight value of each frequency band is obtained through the information entropy calculation formula. When the system detects an abnormal event that exceeds a predetermined threshold (such as the low-frequency energy coefficient of variation >30%), it automatically triggers the weight compensation algorithm to increase the weight coefficient of the corresponding frequency band by 20%-50%. The entropy weight aggregation process adopts time decay coupling technology, that is, the fusion weight of each sliding window is dynamically adjusted according to the time distance. The closer the window is to the current moment, the higher the weight is, forming a weight gradient field that decays exponentially with time, and finally generating a spatiotemporal fusion feature tensor;

[0071] The spatiotemporal fusion feature tensor of each visitor is divided into segments along the time axis, and the coordinate values ​​of each segment in the projection space are fitted with cubic spline interpolation, and finally the dynamic portrait matrix of each visitor is output.

[0072] Platform carrying capacity analysis module: Analyzes the visitor dynamic profile matrix based on a deep spatiotemporal convolutional network to extract the resource demand characteristics of each visitor; and then generates a threat assessment index for the platform's carrying capacity;

[0073] Extract the resource demand characteristics of each visitor. The specific extraction logic is as follows:

[0074] Performing a three-dimensional time window sliding dimensionality reduction on the spatiotemporal synchronization feature tensor based on a deep spatiotemporal convolutional network to obtain spatiotemporal feature dimensionality reduction data of each visitor;

[0075] Based on the spatiotemporal feature dimensionality reduction data, the topological clustering of mobile trajectories is identified to obtain clustered behavior trajectory data of each visitor;

[0076] Dynamically segment the behavior sequence of cluster behavior trajectory data to obtain the spatiotemporal distribution matrix of each visitor's behavior interval;

[0077] Based on the spatiotemporal distribution matrix of behavior intervals, a normal behavior baseline pattern is constructed to obtain the probability data of abnormal behavior deviation for each visitor;

[0078] A hidden Markov model of resource demand is established based on the abnormal behavior deviation probability data to extract resource demand characteristics of each visitor.

[0079] In the embodiment of the present invention, the spatiotemporal synchronization feature tensor is input into the deep spatiotemporal convolutional network. After multiple layers of stacking, the spatiotemporal synchronization feature tensor is subjected to a global average pooling operation, and the feature tensor in each time window is compressed into a vector of fixed length to achieve dimensionality reduction of three-dimensional data; the multi-dimensional feature vector after dimensionality reduction is input into the dynamic time warping algorithm to calculate the morphological similarity between the trajectories of each visitor. In the specific operation, a dynamic matching matrix of the trajectory feature sequence is first established, and the minimum cumulative distance is calculated by finding the optimal curved path between the two feature sequences. For scenarios where the group size exceeds the preset threshold, an adaptive hierarchical clustering method is adopted: in the first stage, the spatial grid is coarsely grouped according to the coordinates of the trajectory starting point, and in the second stage, agglomerative clustering is performed within each coarse group based on the DTW distance, and a distance threshold is set. Finally, the trajectory group label to which each visitor belongs and the center trajectory feature vector of the group are output to form cluster behavior trajectory clustering data. A dynamic sliding window is constructed along the time dimension. For each group's trajectory sequence: 1) a keypoint detection algorithm is used to identify behavioral turning points such as stops, reversals, and accelerations; 2) the continuous trajectory is divided into several behavioral segments based on the turning points; 3) an N×M-dimensional spatiotemporal distribution matrix is ​​constructed, where rows represent time intervals and columns represent spatial regions. The matrix elements represent the frequency of the target behavior occurring within that time period and region. Finally, a weighted average is taken of the matrices for all members of the same group to generate a spatiotemporal distribution matrix for each visitor's behavior interval. Each value in the spatiotemporal distribution matrix is ​​substituted into the corresponding kernel density model to calculate the probability of each behavior unit value deviating from the normal range. The deviation probabilities of all units are then weighted and fused to output the probability of abnormal behavior deviation for each visitor. The real-time normal behavior deviation probability data is input into the model as observations, and the Viterbi algorithm is used to decode the most likely hidden state sequence. Demand features are ultimately extracted based on the decoding results.

[0080] The logic for obtaining cluster behavior trajectory data for each visitor is as follows:

[0081] The spectral clustering algorithm is used to detect the peak value of trajectory density of each visitor's spatiotemporal feature dimensionality reduction data and generate cluster center points;

[0082] The morphological similarity between each trajectory segment and the cluster center is calculated based on the dynamic time warping algorithm to obtain the trajectory membership matrix of each visitor;

[0083] Performing a spatiotemporal coupling analysis on the trajectory membership matrix to generate a group behavior coverage heat map of each visitor;

[0084] The behavioral clustering boundary lines are delineated according to the energy distribution density in the heat map to form clustering data of each visitor's cluster behavior trajectory.

[0085] The present invention uses a spectral clustering algorithm to detect density peaks in preprocessed spatiotemporal feature data. A two-dimensional gridded density distribution map is constructed, and a multi-scale sliding window is used to scan the dwell frequency of each area. Initial cluster centers with significant clustering characteristics are selected by combining local density sorting and relative distance thresholds. A dynamic time warping algorithm is then used to perform morphological matching on individual trajectory segments. A three-dimensional feature sequence comparison matrix of azimuth angle, velocity, and dwell duration is constructed to dynamically plan optimal alignment paths and calculate trajectory similarity metrics. This generates a multi-level membership matrix reflecting the strength of association between visitor behavior and cluster centers. A spatiotemporal coupling analysis is then performed on the data within the membership matrix, and the discrete trajectory matching values ​​are mapped to a unified spatiotemporal coordinate system. A dynamic heatmap is generated by fusing a weighted time decay function with spatial kernel density estimation. Finally, intelligent cluster boundary delineation is implemented based on thermal energy density characteristics. An edge detection algorithm is used to extract gradient mutation regions in the heatmap. This is combined with morphological filtering to eliminate noise interference, dynamically generate cluster boundary polygons, and automatically trigger cluster topology optimization based on the energy density change rate to form clustered behavior trajectory data for each visitor.

[0086] Generate a threat assessment index for platform carrying capacity, including:

[0087] The access path graph is modeled based on the resource demand characteristic parameters of each visitor, and the path node influence factor is calculated using the directed graph betweenness centrality algorithm to generate the access path correlation quantification value of each visitor;

[0088] Based on the resource demand characteristic parameters of each visitor, service call sequence pattern mining is carried out, service combination dependency is calculated, and the service dependency quantification value of each visitor is generated;

[0089] Perform concurrent operation simulation on the clustered behavior trajectory data of each visitor, calculate the critical resource competition probability of each visitor, and generate the concurrent impact risk value of each visitor;

[0090] Based on the entropy weight method, dynamic weight allocation is performed on the visitor's access path correlation quantification value, service dependency quantification value and concurrent impact risk value to obtain a three-dimensional indicator weight matrix;

[0091] The three-dimensional indicator weight matrix is ​​normalized through the indicator fusion function, and the Euclidean distance value of the threat assessment index of each visitor is calculated, and then the threat assessment index of the platform carrying capacity is comprehensively generated.

[0092] An embodiment of the present invention constructs a weighted directed path graph model based on visitor access behavior data, adopts the betweenness centrality algorithm to calculate the influence factor of the path node, and generates a quantitative value of the access path association by multiplying the standardized path length and the residence time factor; secondly, the service call log is analyzed by using the sequence pattern mining technology, and the service combination frequency matrix and dependency relationship map are established. The direct dependency strength and indirect coupling degree between services are calculated based on the conditional probability model and time series correlation to generate a quantitative index of service dependency; at the same time, the concurrent operation scenario under the cluster behavior mode is reconstructed by the Monte Carlo simulation method, and the pressure propagation model is modeled in combination with the resource dependency tree to calculate the concurrent impact risk value; then, the information entropy theory is introduced to dynamically evaluate the discrete degree of the distribution of each indicator data, and the indicator weight is adjusted in real time according to the inverse proportional distribution law of the entropy value. Finally, the weighted Euclidean distance calculation is performed to generate a threat assessment index of the platform carrying capacity.

[0093] Platform carrying capacity warning module: Based on the threat assessment index of the platform carrying capacity, it provides early warning management of the platform's carrying capacity.

[0094] Carry out early warning management of the platform's carrying capacity. The specific operations are as follows:

[0095] Compare the threat assessment index of the platform's carrying capacity with the preset threat assessment threshold range [S1, S2];

[0096] If the threat assessment index of the platform's carrying capacity is less than S1, the platform is judged to be in a low-risk state and the conventional monitoring strategy is implemented;

[0097] If the threat assessment index of the platform's carrying capacity is greater than S2, the platform is judged to be in a high-risk state, triggering an emergency warning response and immediately activating the resource elastic expansion mechanism;

[0098] If the threat assessment index of the platform's carrying capacity is within the threat assessment threshold range, the platform is judged to be in a medium-risk state, and the dynamic load balancing strategy is activated to optimize resource allocation priority.

[0099] It should be noted that the intervals and thresholds are set for ease of comparison. The threshold size depends on the amount of sample data and the cardinality set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless numerical calculations. These formulas are derived from software simulations of the most recent real-world conditions using large amounts of data. The preset parameters in these formulas are set by those skilled in the art based on actual conditions.

[0100] 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.

[0101] It should be understood that determining B based on A does not mean determining B based solely on A. B can also be determined based on A and / or other information.

[0102] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0103] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. Intelligent visitor information management system based on big data analysis, characterized by: include: Visitor portrait generation module: collects multi-dimensional visitor information flow from the platform backend database in real time; Perform multimodal fusion processing on multi-dimensional visitor information flow to build a dynamic portrait matrix for each visitor; Platform carrying capacity analysis module: Analyzes the visitor dynamic profile matrix based on a deep spatiotemporal convolutional network to extract the resource demand characteristics of each visitor; and then generates a threat assessment index for the platform's carrying capacity; Platform carrying capacity warning module: Based on the threat assessment index of the platform carrying capacity, it provides early warning management of the platform's carrying capacity.

2. The visitor information intelligent management system based on big data analysis according to claim 1 is characterized in that: Perform multimodal fusion processing on the multi-dimensional visitor information flow to construct a dynamic portrait matrix for each visitor. The specific construction process is as follows: The multi-dimensional visitor information flow includes the visit behavior data, network protocol data and session trajectory data of each visitor; Performing multi-source spatiotemporal benchmark alignment processing on the access behavior data, network protocol data, and session trajectory data to generate multimodal fusion base data for each visitor; Performing three-dimensional feature decoupling on the multimodal fusion base data to generate visit behavior attribute data, network protocol data, and session trajectory sequence data of each visitor; Based on the access behavior attribute data, cross-modal correlation analysis is performed on the session trajectory sequence data and the network protocol data, dynamic coupling weights are calculated through a bidirectional attention mechanism, and a behavior-protocol-trajectory correlation matrix of each visitor is generated; The behavior-protocol-trajectory association matrix is ​​dynamically fused in time and space to construct a dynamic portrait matrix of each visitor.

3. The visitor information intelligent management system based on big data analysis according to claim 2 is characterized in that: The three-dimensional feature decoupling of the multimodal fusion base data is performed, and the specific logic is as follows: Establish an operation type-frequency association matrix, count the operation type distribution of each visitor according to the preset time window, normalize the original frequency values ​​by standard deviation, assign a weight coefficient of 2 times to high-frequency abnormal operation types, and generate a weighted operation entropy value. Simultaneously, use a sliding window mechanism to construct an operation time series histogram, calculate the coefficient of variation of the proportion of the main operation types within the window, and finally linearly combine the weighted entropy value and the coefficient of variation to output a composite attribute indicator representing the concentration of each visitor's behavior. This is used as the visit behavior attribute data of each visitor. A pattern recognition algorithm is used to analyze the sequence of pages continuously visited by each visitor, and the set of frequent jump paths that meet the minimum support threshold is counted. A probability distribution is fitted for the samples of the dwell time of each page type, and the shape parameter α and scale parameter β are extracted. A state transition matrix between pages is constructed, and the number of jumps from the current interface to the possible next interface {P1, P2, ... Pn} by each visitor is counted. The probability value of each jump direction is calculated to form a discrete probability distribution. Based on the discrete characteristics of the probability distribution, the Shannon entropy formula is applied to calculate its entropy value, which is used to quantify the degree of certainty of each visitor's navigation path, and then the session trajectory sequence data of each visitor is determined.

4. The visitor information intelligent management system based on big data analysis according to claim 2 is characterized in that: The behavior-protocol-trajectory association matrix is ​​subjected to dynamic spatiotemporal fusion processing. The specific processing process is as follows: Preprocessing the correlation matrix within a preset time sliding window to calculate the energy spectrum density of each visitor; Based on the energy spectrum density, entropy weights are dynamically aggregated to generate the spatiotemporal synchronization feature tensor of each visitor; The spatiotemporal synchronization feature tensor is input into the feature projection space to construct a dynamic portrait matrix of each visitor.

5. The visitor information intelligent management system based on big data analysis according to claim 1 is characterized in that: Extract the resource demand characteristics of each visitor. The specific extraction logic is as follows: Performing a three-dimensional time window sliding dimensionality reduction on the spatiotemporal synchronization feature tensor based on a deep spatiotemporal convolutional network to obtain spatiotemporal feature dimensionality reduction data of each visitor; Based on the spatiotemporal feature dimensionality reduction data, the topological clustering of mobile trajectories is identified to obtain clustered behavior trajectory data of each visitor; Dynamically segment the behavior sequence of cluster behavior trajectory data to obtain the spatiotemporal distribution matrix of each visitor's behavior interval; Based on the spatiotemporal distribution matrix of behavior intervals, a normal behavior baseline pattern is constructed to obtain the probability data of abnormal behavior deviation for each visitor; A model is established based on the abnormal behavior deviation probability data to extract the resource demand characteristics of each visitor.

6. The visitor information intelligent management system based on big data analysis according to claim 5 is characterized in that: The logic for obtaining cluster behavior trajectory data for each visitor is as follows: The spectral clustering algorithm is used to detect the peak value of trajectory density of each visitor's spatiotemporal feature dimensionality reduction data and generate cluster center points; The morphological similarity between each trajectory segment and the cluster center is calculated based on the dynamic time warping algorithm to obtain the trajectory membership matrix of each visitor; Performing a spatiotemporal coupling analysis on the trajectory membership matrix to generate a group behavior coverage heat map of each visitor; The behavioral clustering boundary lines are delineated according to the energy distribution density in the heat map to form clustering data of each visitor's cluster behavior trajectory.

7. The visitor information intelligent management system based on big data analysis according to claim 1 is characterized in that: Generate a threat assessment index for platform carrying capacity, including: The access path graph is modeled based on the resource demand characteristic parameters of each visitor, and the path node influence factor is calculated using the directed graph betweenness centrality algorithm to generate the access path correlation quantification value of each visitor; Based on the resource demand characteristic parameters of each visitor, service call sequence pattern mining is carried out, service combination dependency is calculated, and the service dependency quantification value of each visitor is generated; Perform concurrent operation simulation on the clustered behavior trajectory data of each visitor, calculate the critical resource competition probability of each visitor, and generate the concurrent impact risk value of each visitor; Based on the entropy weight method, dynamic weight allocation is performed on the visitor's access path correlation quantification value, service dependency quantification value and concurrent impact risk value to obtain a three-dimensional indicator weight matrix; The three-dimensional indicator weight matrix is ​​normalized through the indicator fusion function, and the Euclidean distance value of the threat assessment index of each visitor is calculated, and then the threat assessment index of the platform carrying capacity is comprehensively generated.

8. The visitor information intelligent management system based on big data analysis according to claim 1 is characterized in that: Carry out early warning management of the platform's carrying capacity. The specific operations are as follows: Compare the threat assessment index of the platform's carrying capacity with the preset threat assessment threshold range [S1, S2]; If the threat assessment index of the platform's carrying capacity is less than S1, the platform is judged to be in a low-risk state and the conventional monitoring strategy is implemented; If the threat assessment index of the platform's carrying capacity is greater than S2, the platform is judged to be in a high-risk state, triggering an emergency warning response and immediately activating the resource elastic expansion mechanism; If the threat assessment index of the platform's carrying capacity is within the threat assessment threshold range, the platform is judged to be in a medium-risk state, and the dynamic load balancing strategy is activated to optimize resource allocation priority.

Images