Artificial intelligence-based cross-border e-commerce data analysis method and device

Abstract

This invention discloses an artificial intelligence-based method and apparatus for cross-border e-commerce data analysis, belonging to the field of data analysis technology. The method includes: real-time acquisition of multi-source heterogeneous data from a cross-border e-commerce platform, preprocessing the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection analysis; constructing a cross-border e-commerce data analysis architecture based on a deep learning algorithm, and introducing an enhanced dung beetle optimization algorithm for optimization to obtain an anomaly detection analysis model; inputting the standardized feature vectors into the anomaly detection analysis model to obtain a comprehensive anomaly score, and performing anomaly judgment and anomaly attribution. This invention solves the problems of low anomaly detection accuracy, difficulty in model parameter optimization, and weak multi-source data fusion capability in existing technologies for cross-border e-commerce.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a cross-border e-commerce data analysis method and apparatus based on artificial intelligence. Background Technology

[0002] With the deepening of global economic integration, the cross-border e-commerce industry has experienced explosive growth. Cross-border e-commerce platforms generate massive amounts of data daily, including product traffic, user behavior, order transactions, logistics tracking, and account operations. This data is characterized by its multi-source heterogeneity, high dimensionality, and strong time-series nature.

[0003] Currently, traditional cross-border e-commerce data analysis methods mainly rely on simple statistical rules or threshold judgments, which are insufficient to cope with complex and ever-changing business scenarios and highly concealed abnormal behaviors (such as order fraud, malicious refunds, and logistics anomalies). Existing machine learning methods typically rely on human experience for feature engineering and model parameter tuning, which is not only inefficient but also prone to getting trapped in local optima, resulting in low detection accuracy. Furthermore, the anomaly detection capabilities of a single model are limited, making it difficult to capture different types of anomaly patterns in the data.

[0004] Therefore, there is an urgent need for a cross-border e-commerce data analysis method that can automatically extract features, adaptively optimize model parameters, and integrate multi-dimensional data for comprehensive analysis. Summary of the Invention

[0005] This invention provides an artificial intelligence-based method and apparatus for cross-border e-commerce data analysis, relating to the field of data analysis technology. This invention addresses the problems of low accuracy in anomaly detection, difficulty in optimizing model parameters, and weak multi-source data fusion capabilities in existing technologies for cross-border e-commerce.

[0006] In a first aspect, embodiments of the present invention provide a cross-border e-commerce data analysis method based on artificial intelligence, the method comprising: Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, and preprocessing of the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection and analysis; Based on deep learning algorithms, a cross-border e-commerce data analysis architecture was constructed, and an enhanced dung beetle optimization algorithm was introduced for optimization to obtain an anomaly detection and analysis model. The standardized feature vector is input into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and then anomaly determination and anomaly attribution are performed.

[0007] The technical solution provided in this application has at least the following beneficial effects: By collecting multi-source heterogeneous data, the operational status of cross-border e-commerce can be captured comprehensively, avoiding the one-sidedness of a single data source. An enhanced dung beetle optimization algorithm is introduced to automatically optimize the hyperparameters of the deep learning architecture. By using chaotic initialization, convergence factor, and PSO mechanism, the problem of traditional algorithms easily getting trapped in local optima is overcome, significantly improving the detection accuracy and generalization ability of the model. An anomaly detection and analysis model is constructed by combining the hybrid architecture of LSTM-AE and IF algorithms. It can not only use deep learning to extract time-series features, but also use statistical methods to capture isolated anomalies. With dynamic weighted fusion, it can achieve accurate identification of complex anomalies. The TreeSHAP algorithm is used for anomaly attribution, which enhances the interpretability of the model and provides operators with intuitive basis for anomaly handling.

[0008] In one optional implementation, multi-source heterogeneous data from a cross-border e-commerce platform is collected in real time, and the multi-source heterogeneous data is preprocessed to obtain a standardized feature vector for anomaly detection analysis, including: Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, including product traffic data, user conversion funnel data, order refund logs, logistics trajectory data, and account operation logs; Standardized cross-border e-commerce data is obtained by preprocessing multi-source heterogeneous data, including time window alignment, missing value imputation and outlier truncation, and standardization. Standardized feature vectors are constructed from standardized cross-border e-commerce data to obtain standardized feature vectors for anomaly detection and analysis.

[0009] In one alternative implementation, a cross-border e-commerce data analysis architecture is constructed based on deep learning algorithms, and an enhanced dung beetle optimization algorithm is introduced for optimization to obtain an anomaly detection and analysis model, including: A data analysis architecture for cross-border e-commerce was built based on deep learning algorithms. The hyperparameters of the cross-border e-commerce data analysis architecture are encoded as the position vectors of individuals in the enhanced dung beetle optimization algorithm, and the accuracy function on the validation set is set as the fitness function. Based on the fitness function, the hyperparameters of the cross-border e-commerce data analysis architecture are optimized using the enhanced dung beetle optimization algorithm to obtain an anomaly detection and analysis model.

[0010] In one optional implementation, the cross-border e-commerce data analysis architecture includes a feature reconstruction module, a statistical anomaly generation module, a reconstruction error calculation module, and a dynamic weighted fusion module. The feature reconstruction module is equipped with a feature reconstruction sub-model, which is constructed based on the LSTM-AE algorithm. The statistical anomaly generation module is equipped with a statistical anomaly generation sub-model, which is constructed based on the IF algorithm.

[0011] In one alternative implementation, based on the fitness function, an enhanced dung beetle optimization algorithm is used to optimize the hyperparameters of the cross-border e-commerce data analysis architecture, resulting in an anomaly detection and analysis model, including: The chaotic sequence is generated using the Logistic mapping and then mapped to the solution space of individuals in the enhanced dung beetle optimization algorithm to obtain an initial population including several initial individuals. Based on the fitness function, calculate the fitness value of each individual in the initial population or the updated population of the previous iteration, and obtain the individual's best position and the global best individual of the population based on the fitness value. According to a preset ratio, the initial population or the population updated in the previous iteration is divided into the current iteration's dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population. By introducing a convergence factor and a PSO mechanism, the corresponding rolling, dancing, breeding, foraging, and stealing behaviors are executed on the current iteration's rolling dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population to obtain the updated population for the current iteration. Repeatedly update the position of the population. When the current iteration reaches the maximum number of iterations or the fitness value of the global best individual meets the requirements, terminate the iterative update of the population and output the final global best individual. The position vector of the final globally optimal individual is decoded to obtain the optimal hyperparameters of the cross-border e-commerce data analysis architecture. The cross-border e-commerce data analysis architecture with the optimal hyperparameters is then trained to obtain the anomaly detection analysis model.

[0012] In one optional implementation, a standardized feature vector is input into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and anomaly determination and attribution are performed, including: The standardized feature vector is input into the anomaly detection and analysis model, and the feature reconstruction sub-model of the optimal feature reconstruction module is used to generate the reconstructed vector of the standardized feature vector. The statistical anomaly generation sub-model of the optimal statistical anomaly generation module is used to generate statistical anomaly scores for standardized feature vectors. The optimal reconstruction error calculation module is used to calculate the Euclidean distance between the reconstructed vector and the standardized feature vector as the reconstruction error. An adaptive fusion weight is introduced, and the optimal dynamic weighted fusion module is used to dynamically weight and fuse the statistical anomaly score and the reconstruction error to obtain the comprehensive anomaly score. Based on the comprehensive anomaly score, anomaly detection is performed using a dynamic adaptive threshold to obtain the anomaly detection results for cross-border e-commerce data analysis. If the anomaly determination result indicates the presence of an anomaly, an anomaly attribution based on interpretability analysis is performed, and a corresponding structured anomaly report is generated.

[0013] In one optional implementation, the formula for calculating the comprehensive anomaly score is: In the formula, The overall abnormal score; This represents the reconstruction error; To statistically analyze outlier scores; This is the normalization function; For adaptive fusion weights.

[0014] In one optional implementation, anomaly detection based on a dynamic adaptive threshold is performed according to the comprehensive anomaly score to obtain the anomaly detection results for cross-border e-commerce data analysis, including: By summing the comprehensive anomaly scores at several time points, a comprehensive anomaly score sequence is obtained; Calculate the moving average and weighted moving standard deviation of the comprehensive anomaly score sequence, and generate the corresponding dynamic adaptive threshold; Compare the current comprehensive anomaly score with the dynamic adaptive threshold. If the score exceeds the dynamic adaptive threshold, an anomaly is determined to exist, and the corresponding anomaly determination result is output. Otherwise, the score is determined to be normal, and the corresponding anomaly determination result is output.

[0015] In one optional implementation, if the anomaly determination result indicates the existence of an anomaly, an anomaly attribution based on interpretability analysis is performed, and a corresponding structured anomaly report is generated, including: The TreeSHAP algorithm is used to calculate the contribution of each feature in the standardized feature vector to the overall anomaly score, resulting in a feature contribution vector. Sort the feature contribution vector in descending order of absolute value, select the top few features in the contribution ranking as the most critical features, and obtain the list of the most critical features that cause the anomaly. Based on the list of most critical features, generate a structured anomaly report that includes anomaly time, anomaly score, anomaly type, core influencing factors, and handling instructions.

[0016] Secondly, embodiments of the present invention provide an artificial intelligence-based cross-border e-commerce data analysis device for implementing a cross-border e-commerce data analysis method, the device comprising: The data acquisition unit is used to collect multi-source heterogeneous data from cross-border e-commerce platforms in real time, and to preprocess the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection and analysis. The model building unit is used to build a cross-border e-commerce data analysis architecture based on deep learning algorithms, and to introduce an enhanced dung beetle optimization algorithm for optimization to obtain an anomaly detection and analysis model. The data analysis unit is used to input standardized feature vectors into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and to determine and attribute anomalies.

[0017] A third aspect of this invention provides an electronic device, which includes: At least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by at least one processor, such that the at least one processor can perform the method proposed in the first aspect of the present invention.

[0018] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method as described in the first aspect of the present invention. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the steps of an artificial intelligence-based cross-border e-commerce data analysis method provided in an embodiment of the present invention. Figure 3 This is a functional unit diagram of an artificial intelligence-based cross-border e-commerce data analysis device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0021] The present invention will be further described below with reference to the accompanying drawings.

[0022] Reference Figure 1 , Figure 1 This is a schematic diagram of the electronic device structure of the hardware operating environment involved in the embodiments of the present invention.

[0023] like Figure 1As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. The memory 1005 may also optionally be a storage device independent of the aforementioned processor 1001.

[0024] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0025] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and an electronic program for an AI-based cross-border e-commerce data analysis device.

[0026] exist Figure 1 In the electronic device shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the electronic device of the present invention can be set in the electronic device. The electronic device calls the electronic program of the artificial intelligence-based cross-border e-commerce data analysis device stored in the memory 1005 through the processor 1001, and executes the artificial intelligence-based cross-border e-commerce data analysis method provided in the embodiment of the present invention.

[0027] Reference Figure 2 The present invention provides an artificial intelligence-based cross-border e-commerce data analysis method, the method comprising: S201: Collect multi-source heterogeneous data from cross-border e-commerce platforms in real time, and preprocess the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection and analysis. S202: Based on deep learning algorithms, a cross-border e-commerce data analysis architecture is constructed, and an enhanced dung beetle optimization algorithm is introduced for optimization to obtain an anomaly detection and analysis model; S203: Input the standardized feature vector into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and then perform anomaly determination and anomaly attribution.

[0028] The technical solution provided in this application has at least the following beneficial effects: By collecting multi-source heterogeneous data, the operational status of cross-border e-commerce can be captured comprehensively, avoiding the one-sidedness of a single data source. An enhanced dung beetle optimization algorithm is introduced to automatically optimize the hyperparameters of the deep learning architecture. Chaotic initialization, convergence factors, and particle swarm optimization (PSO) mechanisms overcome the problem of traditional algorithms easily getting trapped in local optima, significantly improving the model's detection accuracy and generalization ability. A hybrid architecture combining Long Short-Term Memory (LSTM) networks, autoencoders (AE), and Isolation Forest (IF) algorithms is used to construct an anomaly detection and analysis model. This model can extract temporal features using deep learning and capture isolated anomalies using statistical methods. Combined with dynamic weighted fusion, it achieves accurate identification of complex anomalies. The Tree Shapley Additive exPlanations (TreeSHAP) algorithm is used for anomaly attribution, enhancing the model's interpretability and providing operators with intuitive basis for anomaly handling.

[0029] In one optional implementation, multi-source heterogeneous data from a cross-border e-commerce platform is collected in real time, and the multi-source heterogeneous data is preprocessed to obtain a standardized feature vector for anomaly detection analysis, including: S2011: Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, including product traffic data, user conversion funnel data, order refund logs, logistics trajectory data, and account operation logs; Product traffic data includes product page views, unique visitors, average dwell time, bounce rate, etc.; the data is usually stored in the form of log streams. User conversion funnel data: records the user's conversion behavior throughout the entire process from "exposure - click - add to cart - place order - payment", including the conversion rate and drop-off points at each stage; Order refund logs include structured data such as refund initiation time, refund amount, refund reason code, and dispute status. Logistics tracking data includes pickup time, updates to transit milestones, estimated delivery time, actual receipt time, and flags of any abnormal logistics statuses. Account operation logs contain sensitive security information such as user login IP, login device fingerprint, password modification records, and delivery address change records. S2012: Preprocessing of multi-source heterogeneous data by time window alignment, missing value imputation and outlier truncation, and standardization to obtain standardized cross-border e-commerce data; Time window alignment: Set a uniform time window (e.g., 5 minutes or 1 hour per window) to aggregate or interpolate data from different data sources according to the window. For example, map logistics status updates to the most recent time window to ensure that each row of data corresponds to the multidimensional features of the same time period; Missing value imputation and outlier truncation: For missing values ​​of numerical features, impute them using the mean of the feature over the past 24 hours; for categorical features, impute them using the mode or by adding an "unknown" category. Outliers are identified using box plots. For data that exceeds a reasonable range (such as abnormally high page views), truncation is performed by setting upper and lower boundary values ​​to prevent extreme values ​​from interfering with model training. Standardization: Z-Score standardization is used to process the data to eliminate the influence of different feature units; S2013: Standardize the feature vectors of standardized cross-border e-commerce data to obtain standardized feature vectors for anomaly detection and analysis.

[0030] In one alternative implementation, a cross-border e-commerce data analysis architecture is constructed based on deep learning algorithms, and an enhanced dung beetle optimization algorithm is introduced for optimization to obtain an anomaly detection and analysis model, including: S2021: Building a cross-border e-commerce data analysis architecture based on deep learning algorithms; S2022: Encode the hyperparameters of the cross-border e-commerce data analysis architecture into the position vectors of individuals in the enhanced dung beetle optimization algorithm, and set the accuracy function on the validation set as the fitness function; Hyperparameters can include: The number of LSTM layers and the number of LSTM neurons, the dimension of the latent vectors, the learning rate, and the batch size of the encoder and decoder of the feature reconstruction sub-model; The subsampling size, number of trees, and anomaly ratio threshold of the statistical anomaly generation sub-model are considered. Reconstruction error weights and statistical anomaly score weights in the dynamic weighted fusion module; S2023: Based on the fitness function, the hyperparameters of the cross-border e-commerce data analysis architecture are optimized using the enhanced dung beetle optimization algorithm to obtain an anomaly detection and analysis model.

[0031] In one optional implementation, the cross-border e-commerce data analysis architecture includes a feature reconstruction module, a statistical anomaly generation module, a reconstruction error calculation module, and a dynamic weighted fusion module. The feature reconstruction module is equipped with a feature reconstruction sub-model, which is constructed based on the LSTM-AE algorithm. The statistical anomaly generation module is equipped with a statistical anomaly generation sub-model, which is constructed based on the IF algorithm.

[0032] In one alternative implementation, based on the fitness function, an enhanced dung beetle optimization algorithm is used to optimize the hyperparameters of the cross-border e-commerce data analysis architecture, resulting in an anomaly detection and analysis model, including: S20231: Use Logistic mapping to generate chaotic sequences, and map the chaotic sequences to the solution space of individuals in the enhanced dung beetle optimization algorithm to obtain an initial population including several initial individuals; The formula is: In the formula, For the first n+ 1. n There are several chaotic variables whose values ​​range from [0, 1]. The stability coefficient is typically 4. This sequence is ergodic and random, ensuring that the initial population is uniformly distributed in the solution space, avoiding getting trapped in local optima, which is superior to traditional random initialization. n Indicator of chaotic variables; In the formula, For the initial population, the first i An initial individual; For the first i One chaotic variable; These are the upper and lower bounds of the parameter space; i For individual indicators; t This is an indicator of the number of iterations. In the formula, For the initial population, the first i The initial velocity of each initial individual; A random number in the interval (0,1); This represents the maximum speed. S20232: Based on the fitness function, calculate the fitness value of each individual in the initial population or the population updated in the previous iteration, and obtain the individual's best position and the global best individual of the population based on the fitness value. S20233: According to a preset ratio, divide the initial population or the population updated in the previous iteration to obtain the current iteration's dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population. The initial population or the population updated in the previous iteration is sorted in descending order according to the fitness value, and divided into the current iteration's dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population in a ratio of 1:1:1:2. S20234: Introduce convergence factor and PSO mechanism to perform corresponding rolling, dancing, breeding, foraging and stealing behaviors on the current iteration of the rolling dung beetle population, the breeding dung beetle population, the foraging dung beetle population and the stealing dung beetle population to obtain the updated population of the current iteration. generate , Given a random number in the interval [0,1), if If the dung beetle population is unobstructed, the corresponding dung beetles will perform the rolling behavior, as shown in the formula: In the formula, For the first t+ The result of performing the rolling ball behavior in the first iteration is the first... i A newer rolling dung beetle; For the first t The iteration of the ... i A newer dung beetle, in the first iteration, is the initial individual; A random number in the interval [0,1]. For the first t+ The first iteration i The update speed of the newest rolling dung beetle; For the first t The iteration of the ... i The speed of the updated rolling dung beetle, in the first iteration, The initial velocity; Accelerate one's own cognition; The acceleration coefficient of social cognition; For the first t The globally optimal individual in the next iteration; The inertia coefficients are the convergence factor scheme. For the first t The iteration of the ... i The optimal position for an individual dung beetle; In the formula, This represents the maximum number of iterations. These are the maximum and minimum values ​​of the convergence factor; like If a dung beetle in the population encounters an obstacle, it will be designated as a dancing dung beetle and will perform a dancing behavior. The formula is as follows: In the formula, Number of iterations t+ The first result obtained by performing the dance act i A newer dancing dung beetle; It is the tangent function; The parameters for the dancing angle determine the intensity and direction of the new direction; Number of iterations t The first time encountering obstacles i The rolling dung beetle, also known as the dancing dung beetle; The formula for implementing reproductive behavior in a reproductive dung beetle population is as follows: In the formula, Number of iterations t+ 1. The first result obtained from performing reproductive behavior i A new breeding dung beetle; Number of iterations t The first result obtained from performing reproductive behavior i A new breeding dung beetle; A random number in the interval [0,1]. The formula for observing foraging behavior in a foraging dung beetle population is as follows: In the formula, Number of iterations t+ 1. The first result obtained from performing foraging behavior i A newer foraging dung beetle; Number of iterations t The i A newer foraging dung beetle; This is the step size control factor, usually set to 0.1; It is the XOR symbol; The symbol for tensor product; The formula for executing theft behavior by dung beetles within a dung beetle population is as follows: In the formula, Number of iterations t+ 1. The first result obtained from the act of theft i A newer dung beetle; Number of iterations t The iA newer dung beetle; For the first t The iteration of the ... i The optimal position for an individual dung beetle that steals dung; These are random numbers distributed according to a standard normal distribution. This is the disturbance intensity coefficient; S20235: Repeatedly update the position of the population. When the current iteration reaches the maximum number of iterations or the fitness value of the global best individual meets the requirements, terminate the iterative update of the population and output the final global best individual. S20236: Decode the position vector of the final globally optimal individual to obtain the optimal hyperparameters of the cross-border e-commerce data analysis architecture, and train the cross-border e-commerce data analysis architecture with the optimal hyperparameters to obtain the anomaly detection analysis model.

[0033] In one optional implementation, a standardized feature vector is input into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and anomaly determination and attribution are performed, including: S2031: Input the standardized feature vector into the anomaly detection analysis model, and use the feature reconstruction sub-model of the optimal feature reconstruction module to generate the reconstructed vector of the standardized feature vector; The feature reconstruction sub-model includes an encoder and a decoder; Encoder: After two LSTM layers, the temporal features of the normalized feature vector are extracted, and finally mapped to a low-dimensional latent vector through a fully connected layer; Decoder: The latent vector is restored to its temporal length through a repetition layer, and then passed through two LSTM layers and a fully connected output layer to generate a reconstructed vector; S2032: The statistical anomaly generation sub-model of the optimal statistical anomaly generation module generates statistical anomaly scores for standardized feature vectors. S2033: Use the optimal reconstruction error calculation module to calculate the Euclidean distance between the reconstructed vector and the standardized feature vector as the reconstruction error; S2034: Introduce adaptive fusion weights and use the optimal dynamic weighted fusion module to dynamically weight and fuse statistical anomaly scores and reconstruction errors to obtain a comprehensive anomaly score; S2035: Based on the comprehensive anomaly score, perform anomaly judgment based on dynamic adaptive threshold to obtain the anomaly judgment result of cross-border e-commerce data analysis; S2036: If the anomaly determination result indicates the existence of an anomaly, then perform anomaly attribution based on interpretability analysis and generate a corresponding structured anomaly report.

[0034] In one optional implementation, the formula for calculating the comprehensive anomaly score is: In the formula, The overall abnormal score; This represents the reconstruction error; To statistically analyze outlier scores; This is the normalization function; For adaptive fusion weights.

[0035] In one optional implementation, anomaly detection based on a dynamic adaptive threshold is performed according to the comprehensive anomaly score to obtain the anomaly detection results for cross-border e-commerce data analysis, including: S20351: Summarize the comprehensive anomaly scores at several time points to obtain the comprehensive anomaly score sequence; S20352: Calculate the moving average and weighted moving standard deviation of the comprehensive anomaly score sequence, and generate the corresponding dynamic adaptive threshold. The formula is as follows: In the formula, For dynamic adaptive thresholds; It is a moving average; The weighted moving standard deviation; This is the sensitivity coefficient; This is the function for calculating standard deviation; This is a comprehensive anomaly score sequence; It is a fluctuation attenuation factor; S20353: Compare the current comprehensive anomaly score with the dynamic adaptive threshold. If it exceeds the dynamic adaptive threshold, it is determined that there is an anomaly and the corresponding anomaly judgment result is output. Otherwise, it is determined that it is normal and the corresponding anomaly judgment result is output.

[0036] In one optional implementation, if the anomaly determination result indicates the existence of an anomaly, an anomaly attribution based on interpretability analysis is performed, and a corresponding structured anomaly report is generated, including: S20361: Using the TreeSHAP algorithm, calculate the contribution of each feature in the standardized feature vector to the overall anomaly score, obtaining the feature contribution vector. The formula is: In the formula, The overall abnormal score; The baseline value for comprehensive anomaly scores is denoted as , and the average anomaly score is denoted as . For the first The contribution of features to the overall anomaly score; For characteristic indicators; The number of features; S20362: Sort the feature contribution in the feature contribution vector in descending order of absolute value, select the top few features in the contribution ranking as the most critical features, and obtain the list of the most critical features that cause the anomaly. S20363: Generate a structured anomaly report based on the list of most critical features, including anomaly time, anomaly score, anomaly type, core influencing factors, and handling instructions.

[0037] This invention also provides an artificial intelligence-based cross-border e-commerce data analysis device 300, with reference to... Figure 3 The device may include the following units: The data acquisition unit 301 is used to collect multi-source heterogeneous data from the cross-border e-commerce platform in real time, and to preprocess the multi-source heterogeneous data to obtain a standardized feature vector for anomaly detection and analysis. Model building unit 302 is used to build a cross-border e-commerce data analysis architecture based on deep learning algorithms, and to introduce an enhanced dung beetle optimization algorithm for optimization to obtain an anomaly detection and analysis model; The data analysis unit 303 is used to input standardized feature vectors into the anomaly detection analysis model to obtain a comprehensive anomaly score, and to perform anomaly determination and anomaly attribution.

[0038] Based on the same inventive concept, another embodiment of the present invention provides an electronic device, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus. Memory, used to store computer programs; The processor, when executing a program stored in memory, implements the AI-based cross-border e-commerce data analysis method of the present invention.

[0039] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EI) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of representation, only one thick line is used in the diagram, but this does not indicate that there is only one bus or one type of bus. The communication interface is used for communication between the aforementioned terminal and other devices. The memory can include Random Access Memory (RAM), or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor.

[0040] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0041] Furthermore, to achieve the above objectives, embodiments of the present invention also propose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the artificial intelligence-based cross-border e-commerce data analysis method of the embodiments of the present invention.

[0042] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable hardware devices (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] The embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0044] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0045] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0046] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. "And / or" indicates that either one or both can be chosen. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes the element.

[0047] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A data analysis method for cross-border e-commerce based on artificial intelligence, characterized in that, The method includes: Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, and preprocessing of the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection and analysis; Based on deep learning algorithms, a cross-border e-commerce data analysis architecture was constructed, and an enhanced dung beetle optimization algorithm was introduced for optimization to obtain an anomaly detection and analysis model. The standardized feature vector is input into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and then anomaly determination and anomaly attribution are performed.

2. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 1, characterized in that, Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, followed by preprocessing of the data to obtain standardized feature vectors for anomaly detection and analysis, including: Real-time collection of multi-source heterogeneous data from cross-border e-commerce platforms, including product traffic data, user conversion funnel data, order refund logs, logistics trajectory data, and account operation logs; Standardized cross-border e-commerce data is obtained by preprocessing multi-source heterogeneous data, including time window alignment, missing value imputation and outlier truncation, and standardization. Standardized feature vectors are constructed from standardized cross-border e-commerce data to obtain standardized feature vectors for anomaly detection and analysis.

3. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 2, characterized in that, Based on deep learning algorithms, a cross-border e-commerce data analysis architecture is constructed, and an enhanced dung beetle optimization algorithm is introduced for optimization, resulting in an anomaly detection and analysis model, including: A data analysis architecture for cross-border e-commerce was built based on deep learning algorithms. The hyperparameters of the cross-border e-commerce data analysis architecture are encoded as the position vectors of individuals in the enhanced dung beetle optimization algorithm, and the accuracy function on the validation set is set as the fitness function. Based on the fitness function, the hyperparameters of the cross-border e-commerce data analysis architecture are optimized using the enhanced dung beetle optimization algorithm to obtain an anomaly detection and analysis model.

4. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 3, characterized in that, The cross-border e-commerce data analysis architecture includes a feature reconstruction module, a statistical anomaly generation module, a reconstruction error calculation module, and a dynamic weighted fusion module. The feature reconstruction module is equipped with a feature reconstruction sub-model, which is constructed based on the LSTM-AE algorithm. The statistical anomaly generation module is equipped with a statistical anomaly generation sub-model, which is constructed based on the IF algorithm.

5. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 4, characterized in that, Based on the fitness function, the hyperparameters of the cross-border e-commerce data analysis architecture are optimized using the enhanced dung beetle optimization algorithm, resulting in an anomaly detection and analysis model, including: The chaotic sequence is generated using the Logistic mapping and then mapped to the solution space of individuals in the enhanced dung beetle optimization algorithm to obtain an initial population including several initial individuals. Based on the fitness function, calculate the fitness value of each individual in the initial population or the updated population of the previous iteration, and obtain the individual's best position and the global best individual of the population based on the fitness value. According to a preset ratio, the initial population or the population updated in the previous iteration is divided into the current iteration's dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population. By introducing a convergence factor and a PSO mechanism, the corresponding rolling, dancing, breeding, foraging, and stealing behaviors are executed on the current iteration's rolling dung beetle population, breeding dung beetle population, foraging dung beetle population, and stealing dung beetle population to obtain the updated population for the current iteration. Repeatedly update the position of the population. When the current iteration reaches the maximum number of iterations or the fitness value of the global best individual meets the requirements, terminate the iterative update of the population and output the final global best individual. The position vector of the final globally optimal individual is decoded to obtain the optimal hyperparameters of the cross-border e-commerce data analysis architecture. The cross-border e-commerce data analysis architecture with the optimal hyperparameters is then trained to obtain the anomaly detection analysis model.

6. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 5, characterized in that, The standardized feature vector is input into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and anomaly determination and attribution are performed, including: The standardized feature vector is input into the anomaly detection and analysis model, and the feature reconstruction sub-model of the optimal feature reconstruction module is used to generate the reconstructed vector of the standardized feature vector. The statistical anomaly generation sub-model of the optimal statistical anomaly generation module is used to generate statistical anomaly scores for standardized feature vectors. The optimal reconstruction error calculation module is used to calculate the Euclidean distance between the reconstructed vector and the standardized feature vector as the reconstruction error. An adaptive fusion weight is introduced, and the optimal dynamic weighted fusion module is used to dynamically weight and fuse the statistical anomaly score and the reconstruction error to obtain the comprehensive anomaly score. Based on the comprehensive anomaly score, anomaly detection is performed using a dynamic adaptive threshold to obtain the anomaly detection results for cross-border e-commerce data analysis. If the anomaly determination result indicates the presence of an anomaly, an anomaly attribution based on interpretability analysis is performed, and a corresponding structured anomaly report is generated.

7. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 6, characterized in that, The formula for calculating the comprehensive anomaly score is as follows: In the formula, The overall abnormal score; This represents the reconstruction error; To statistically analyze outlier scores; This is the normalization function; For adaptive fusion weights.

8. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 7, characterized in that, Based on the comprehensive anomaly score, anomaly detection is performed using a dynamically adaptive threshold, resulting in anomaly detection results for cross-border e-commerce data analysis, including: By summing the comprehensive anomaly scores at several time points, a comprehensive anomaly score sequence is obtained; Calculate the moving average and weighted moving standard deviation of the comprehensive anomaly score sequence, and generate the corresponding dynamic adaptive threshold; Compare the current comprehensive anomaly score with the dynamic adaptive threshold. If the score exceeds the dynamic adaptive threshold, an anomaly is determined to exist, and the corresponding anomaly determination result is output. Otherwise, the score is determined to be normal, and the corresponding anomaly determination result is output.

9. The cross-border e-commerce data analysis method based on artificial intelligence according to claim 8, characterized in that, If the anomaly determination result indicates the presence of an anomaly, an anomaly attribution based on interpretability analysis is performed, and a corresponding structured anomaly report is generated, including: The TreeSHAP algorithm is used to calculate the contribution of each feature in the standardized feature vector to the overall anomaly score, resulting in a feature contribution vector. Sort the feature contribution vector in descending order of absolute value, select the top few features in the contribution ranking as the most critical features, and obtain the list of the most critical features that cause the anomaly. Based on the list of most critical features, generate a structured anomaly report that includes anomaly time, anomaly score, anomaly type, core influencing factors, and handling instructions.

10. An artificial intelligence-based cross-border e-commerce data analysis device, used to implement the cross-border e-commerce data analysis method as described in any one of claims 1-9, characterized in that, The device includes: The data acquisition unit is used to collect multi-source heterogeneous data from cross-border e-commerce platforms in real time, and to preprocess the multi-source heterogeneous data to obtain standardized feature vectors for anomaly detection and analysis. The model building unit is used to build a cross-border e-commerce data analysis architecture based on deep learning algorithms, and to introduce an enhanced dung beetle optimization algorithm for optimization to obtain an anomaly detection and analysis model. The data analysis unit is used to input standardized feature vectors into the anomaly detection and analysis model to obtain a comprehensive anomaly score, and to determine and attribute anomalies.

Images