Cross-border e-commerce platform customer value evaluation method based on artificial intelligence

Abstract

The invention discloses a cross-border e-commerce platform customer value evaluation method based on artificial intelligence, and relates to the technical field of customer value evaluation.A digital twinborn technology is adopted, customer behavior time series data and multi-modal features are fused into a virtual image through nonlinear mapping, discrete data are converted into continuous images to be displayed, and the customer value is evaluated. Therefore, the customer behavior evolution process is visually presented; after time series data mean value aggregation, mapping the time series data mean value with multi-modal features, adjusting influence of each piece of data by using a weight matrix and bias, realizing effective connection between the data, and intuitively displaying a corresponding relationship between a customer state and a behavior; customer behavior data are encoded and decoded through a depth auto-encoder, reconstruction errors are calculated and are converted into abnormal scores through a score mapping function, and abnormal detection is carried out in combination with an isolated forest algorithm, so that early warning signals are generated, and the risk of sharp increase of returned goods and abnormal transaction is recognized in time; the defects of a traditional method in the aspect of real-time risk monitoring are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of customer value evaluation, and in particular to a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence. Background Art

[0002] Currently, for the way of evaluating the value of customers on cross-border e-commerce platforms, most of them construct customer value models based on historical orders, evaluations, and logistics data, and use fixed rules for level / hierarchy division.

[0003] However, in the actual operation process, the information collected by the platform comes from multiple countries and channels, the data is not updated in a timely manner, and the formats are not unified, so it is impossible to effectively capture the instantaneous changes in customer behavior. In international trade, there are significant differences in languages, cultures, and payment habits among different countries, which is likely to result in inaccurate customer portraits. In addition, when the platform faces sudden regional events, holidays, etc., which cause a sharp increase in order returns or abnormal transactions, it can only statistically discover problems afterwards, lacking real-time monitoring and early warning. In response to this, some platforms have set dynamic thresholds and adjustment rules, but still it is difficult to cope with the continuously changing risk patterns in the market environment. There are lags and noises in data collection, resulting in the risk identification model being less sensitive to abnormal behaviors, affecting the accuracy of customer value evaluation. At the same time, static models cannot dynamically track the entire life cycle of customers, so when customer behaviors fluctuate, it is very difficult for the platform to intervene in advance to adjust operation strategies.

[0004] This situation not only weakens the platform's prevention and control capabilities against customer churn risks and fraud behaviors, but also easily misses sudden order returns or abnormal payment behaviors. There is a certain probability that such sudden order returns or abnormal payments have fraud risks or system problems, but the value evaluation process only regards them as low-value or invalid data. In fact, this part of the data has been evaluated and analyzed, but it is actually not effectively identified. It can be seen that the traditional evaluation method has obvious shortcomings in risk early warning. Therefore, there is an urgent need for a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence to solve such problems. Summary of the Invention

[0005] In view of the above existing problems, the present invention is proposed.

[0006] The present invention provides a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence to solve the problems that the traditional value evaluation method has limited analysis of cross-border return rates and abnormal transactions, cannot identify return anomalies in a timely manner, and cannot discover such risks in a timely manner.

[0007] To solve the above technical problems, the present invention provides the following technical solutions:

[0008] An embodiment of the present invention provides a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence, which includes,

[0009] Step S1, collect order, evaluation, return, and channel information, and perform standardization processing on the collected data to obtain preprocessed data;

[0010] Step S2, use natural language processing, image recognition, and speech analysis technologies to process the preprocessed data to extract multimodal features, and use a deep learning model to generate dynamic customer labels;

[0011] Step S3, perform time series analysis on the dynamic customer labels to obtain customer behavior time series data;

[0012] Step S4, use digital twin technology to perform virtual mapping on the customer behavior time series data and multimodal features to form a customer digital twin image;

[0013] Step S5, input the customer behavior time series data into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors, and form a warning signal;

[0014] Step S6, update the customer value evaluation index according to the customer behavior time series data and the warning signal, and stratify the customers.

[0015] As a preferred solution of the customer value evaluation method for a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in Step S2, the steps of using natural language processing, image recognition, and speech analysis technologies to process the preprocessed data to extract multimodal features and using a deep learning model to generate dynamic customer labels are as follows:

[0016] Based on the preprocessed data, adopt an embedding method for text, image, and speech respectively, calculate the multimodal feature vector, and the defined formula is:

[0017] F = σ(W1e1 + W2e2 + W3e3 + b),

[0018] wherein, F represents the multimodal feature vector, e1 represents the text embedding vector, e2 represents the image embedding vector, e3 represents the speech embedding vector, W1 represents the weight matrix corresponding to the text embedding, W2 represents the weight matrix corresponding to the image embedding, W3 represents the weight matrix corresponding to the speech embedding, b represents the bias vector, and σ represents the activation function;

[0019] Use a fully connected layer to map the multimodal features to the label space, defined as:

[0020] z = UF + c, L = softmax(z),

[0021] Among them, z represents the output vector of the fully connected layer, U represents the mapping weight matrix, F represents the aforementioned multi-modal feature vector, c represents the mapping bias vector, L represents the dynamic customer label vector, and softmax represents the normalization function.

[0022] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: the step of extracting the multi-modal features further includes processing social interaction data to generate dynamic customer labels including information on region, culture, and payment habits.

[0023] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: the digital twin image is used to reflect the correspondence between customer behavior time-series data and multi-modal features.

[0024] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in step S4, the step of using digital twin technology to perform virtual mapping on the customer behavior time-series data and multi-modal features to form a customer digital twin image is as follows:

[0025] Perform mean aggregation on the customer behavior time-series data {t1, t2,..., t n}, and the calculation formula is:

[0026]

[0027] Among them, T′ represents the aggregated time-series data vector, t i represents the feature vector at the i-th moment, and n represents the number of moments;

[0028] Fuse the aggregated time-series data vector T ′ with the multi-modal feature vector F, and construct a customer digital twin image through a non-linear mapping function. The formula is:

[0029] D = φ(W t T′ + W f F + d),

[0030] Among them, D represents the customer digital twin image, W t represents the time-series data conversion weight matrix, T ′ represents the aggregated time-series data vector, W f represents the multi-modal feature conversion weight matrix, F represents the multi-modal feature vector, d represents the bias vector, and φ represents the non-linear mapping function.

[0031] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in the anomaly detection step of step S5, an isolation forest algorithm and a deep autoencoder are combined for detection.

[0032] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in step S5, the step of inputting the customer behavior time series data into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate a warning signal is as follows:

[0033] For the customer behavior time series data x, or its aggregated representation, a deep autoencoder is used for encoding, and the encoding process is defined as:

[0034] h = f(x) = σ(W e x + b e ),

[0035] wherein, h represents the encoded feature vector, f(x) represents the encoding function, x represents the input customer behavior time series data vector, W e represents the encoding weight matrix, b e represents the encoding bias vector, and σ represents the activation function;

[0036] The decoding process reconstructs the encoded vector h into an output vector, and the formula is:

[0037]

[0038] wherein, represents the reconstructed vector, g(h) represents the decoding function, W d represents the decoding weight matrix, b d represents the decoding bias vector, and σ represents the activation function.

[0039] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in step S5, the step of inputting the customer behavior time series data into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate a warning signal further includes:

[0040] The reconstruction error is calculated using the Euclidean distance, and the formula is:

[0041]

[0042] wherein, E represents the reconstruction error, and ‖·‖2 represents the Euclidean distance;

[0043] The reconstruction error E is mapped to an anomaly score, and the formula is:

[0044] s = ψ(E),

[0045] where s represents the anomaly score and ψ represents the scoring mapping function;

[0046] After setting the threshold τ, the anomaly score is converted into a warning signal by using the indicator function, and the formula is

[0047] P = 1 {s>τ} ,

[0048] where P represents the warning signal, 1 {s>τ} means taking 1 when s is greater than the threshold τ, otherwise taking 0, and τ represents the detection threshold.

[0049] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in step S6, the step of updating the customer value evaluation index according to the customer behavior time series data and the warning signal and stratifying the customers is

[0050] Combining the historical customer value index V0 with the current behavior time series data aggregation vector T agg and the warning signal P, and calculating the updated customer value index vector V by using the weighted update formula:

[0051] V = ηV0+(1 - η)(T agg +δP),

[0052] where V represents the updated customer value index vector, V0 represents the previous customer value index vector, η represents the smoothing coefficient, T agg represents the behavior time series data aggregation vector, δ represents the warning signal weight, and P represents the warning signal vector.

[0053] As a preferred solution of the method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to the present invention, wherein: in step S6, the step of updating the customer value evaluation index according to the customer behavior time series data and the warning signal and stratifying the customers further includes:

[0054] Stratifying the index vectors of all customers, and using the clustering method to solve the grouping, and the mathematical expression is:

[0055]

[0056] where V i represents the value index vector of the i-th customer, μ k(i) represents the clustering center to which the i-th customer belongs, N represents the total number of customers, and |·| 2 represents the squared Euclidean distance.

[0057] The beneficial effects of the present invention are as follows: The present invention adopts digital twin technology to fuse customer behavior time-series data and multi-modal features into virtual images through non-linear mapping, converting discrete data into continuous image displays, thereby intuitively presenting the evolution process of customer behavior; aggregating the mean of time-series data and jointly mapping it with multi-modal features, adjusting the influence of each data using a weight matrix and bias, and realizing effective docking between data, and intuitively displaying the corresponding relationship between customer status and behavior.

[0058] The present invention encodes and decodes customer behavior data through a deep autoencoder, calculates the reconstruction error, converts it into an anomaly score through a scoring mapping function, and then combines it with the isolation forest algorithm for anomaly detection, thereby generating a warning signal to timely identify the risks of a surge in returns and abnormal transactions; making up for the deficiencies of traditional methods in real-time risk monitoring.

[0059] In addition, the present invention uses a weighted update formula to fuse historical customer value indicators, current time-series data, and warning signals, and then uses a clustering algorithm to stratify customers, realizing the dynamic update and scientific division of customer value indicators.

[0060] In summary, the present invention solves problems such as multi-channel data integration, insufficient expression of customer multi-dimensional features, and lag in risk warning, and has the characteristics of unified parameter definition and mathematical model description, providing a scientific and continuous customer stratification management method for cross-border e-commerce platforms. Brief Description of the Drawings

[0061] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0062] Figure 1 It is a schematic flow chart of a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence of the present invention. Detailed Embodiments

[0063] To make the above objects, features, and advantages of the present invention more obvious and understandable, the following will describe the detailed embodiments of the present invention in conjunction with the drawings of the specification.

[0064] Many specific details are set forth in the following description to facilitate a thorough understanding of the present invention. However, the present invention can also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0065] Second, the so-called "one embodiment" or "embodiment" herein refers to specific features, structures or characteristics that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that are mutually exclusive with other embodiments.

[0066] Embodiment 1, referring to Figure 1 , this embodiment provides a method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence, including the following steps:

[0067] Step S1, collect order, evaluation, return and channel information, and perform standardization processing on the collected data to obtain preprocessed data;

[0068] Step S2, use natural language processing, image recognition and speech analysis technologies to process the preprocessed data to extract multi-modal features, and use a deep learning model to generate dynamic customer labels;

[0069] In step S2, the steps of using natural language processing, image recognition and speech analysis technologies to process the preprocessed data to extract multi-modal features and using a deep learning model to generate dynamic customer labels are as follows:

[0070] Based on the preprocessed data, embedding methods are respectively used for text, image and speech to calculate multi-modal feature vectors, and the defined formula is:

[0071] F = σ(W1e1 + W2e2 + W3e3 + b),

[0072] where F represents the multi-modal feature vector, e1 represents the text embedding vector, e2 represents the image embedding vector, e3 represents the speech embedding vector, W1 represents the weight matrix corresponding to the text embedding, W2 represents the weight matrix corresponding to the image embedding, W3 represents the weight matrix corresponding to the speech embedding, b represents the bias vector, and σ represents the activation function;

[0073] Use the fully connected layer to map the multi-modal features to the label space, defined as:

[0074] z = UF + c, L = softmax(z),

[0075] where z represents the output vector of the fully connected layer, U represents the mapping weight matrix, F represents the aforementioned multi-modal feature vector, c represents the mapping bias vector, L represents the dynamic customer label vector, and softmax represents the normalization function;

[0076] Specifically, here, text, image, and speech feature embeddings are respectively performed on the preprocessed data, and a unified multi-modal feature vector is formed by using linear combination plus activation function. Then, the fully connected layer is used to map this feature vector to the label space, and the dynamic customer labels are generated through the normalization function. This process integrates multi-modal information, enabling the labels to reflect the multi-dimensional behavioral characteristics of customers in cross-border e-commerce, thereby generating dynamic labels;

[0077] The extraction steps of multi-modal features also include processing social interaction data to generate dynamic customer labels containing regional, cultural, and payment habit information;

[0078] Step S3: Perform time series analysis on the dynamic customer labels to obtain customer behavior time series data;

[0079] Step S4: Use digital twin technology to perform virtual mapping on the customer behavior time series data and multi-modal features to form a customer digital twin image;

[0080] The digital twin image is used to reflect the correspondence between the customer behavior time series data and the multi-modal features;

[0081] In step S4, the steps of using digital twin technology to perform virtual mapping on the customer behavior time series data and multi-modal features to form a customer digital twin image are as follows:

[0082] Perform mean aggregation on the customer behavior time series data {t1, t2, …, t n}, and the calculation formula is:

[0083]

[0084] where, T ′ represents the aggregated time series data vector, t i represents the feature vector at the i-th moment, and n represents the number of moments;

[0085] Fuse the aggregated time series data vector T ′ with the multi-modal feature vector F, and construct the customer digital twin image through a non-linear mapping function. The formula is:

[0086] D = φ(W t T ′ + W f F + d),

[0087] where, D represents the customer digital twin image, W t represents the time series data conversion weight matrix, T ′ represents the aggregated time series data vector, W f represents the multi-modal feature conversion weight matrix, F represents the multi-modal feature vector, d represents the bias vector, and φ represents the non-linear mapping function;

[0088] Specifically, the sequential data of customer behavior is aggregated by mean here to obtain a representative vector of the overall sequential features. Then, this aggregated vector and the multi-modal feature vectors extracted in step S2 are jointly input into a non-linear mapping function. The influence of different data is adjusted through a weight matrix and a bias to form a customer digital twin image, which reflects the mapping relationship between the customer's sequential behavior and multi-modal features;

[0089] Adopting the digital twin method helps to transform discrete sequential data and multi-modal information into continuous image representations, facilitating the observation of customer behavior changes;

[0090] Step S5: Input the sequential data of customer behavior into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate warning signals;

[0091] The anomaly detection step in step S5 is performed by combining the isolation forest algorithm and the deep autoencoder;

[0092] In step S5, the step of inputting the sequential data of customer behavior into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate warning signals is as follows:

[0093] For the sequential data x of customer behavior, or its aggregated representation, a deep autoencoder is used for encoding, and the encoding process is defined as:

[0094] h = f(x) = σ(W e x + b e ),

[0095] where h represents the encoded feature vector, f(x) represents the encoding function, x represents the input sequential data vector of customer behavior, W e represents the encoding weight matrix, b e represents the encoding bias vector, and σ represents the activation function;

[0096] The decoding process reconstructs the encoded vector h into an output vector, and the formula is:

[0097]

[0098] where represents the reconstructed vector, g(h) represents the decoding function, W d represents the decoding weight matrix, b d represents the decoding bias vector, and σ represents the activation function;

[0099] In step S5, the step of inputting the customer behavior time series data into the deep autoencoder and the anomaly detection algorithm to detect abnormal returns and fraud risk behaviors and generate warning signals further includes:

[0100] The reconstruction error is calculated using the Euclidean distance, and the formula is:

[0101]

[0102] where E represents the reconstruction error, and ‖·‖2 represents the Euclidean distance;

[0103] Mapping the reconstruction error E to an anomaly score, the formula is:

[0104] s = ψ(E),

[0105] where s represents the anomaly score, and ψ represents the score mapping function;

[0106] After setting the threshold τ, using the indicator function to convert the anomaly score into a warning signal, the formula is

[0107] p = 1 {s>τ} ,

[0108] where P represents the warning signal, 1 {s>τ} means taking 1 when s is greater than the threshold τ, otherwise taking 0, and τ represents the detection threshold;

[0109] Specifically, here the input customer behavior time series data is encoded and decoded by the deep autoencoder to generate a reconstruction vector, and the reconstruction error is calculated using the Euclidean distance. The reconstruction error is used as the basis for anomaly detection, and is converted into an anomaly score through the score mapping function; using the set threshold and the indicator function, the anomaly score is binarized to generate a warning signal, which is used to identify abnormal returns and fraud risk behaviors, and the data anomaly situation is reflected through the reconstruction error;

[0110] Step S6, updating the customer value evaluation index according to the customer behavior time series data and the warning signal, and stratifying the customers;

[0111] In step S6, the step of updating the customer value evaluation index according to the customer behavior time series data and the warning signal and stratifying the customers is,

[0112] Combining the historical customer value index V0 with the current behavior time series data aggregation vector T agg and the warning signal p, using the weighted update formula to calculate the updated customer value index vector V:

[0113] V = ηV0 + (1 - η)(T agg + δP),

[0114] Among them, V represents the updated customer value metric vector, V0 represents the previous customer value metric vector, η represents the smoothing coefficient, T agg represents the aggregated vector of behavioral time-series data, δ represents the early warning signal weight, and P represents the early warning signal vector;

[0115] In step S6, the steps of updating the customer value evaluation metrics based on the customer behavioral time-series data and early warning signals and stratifying the customers further include:

[0116] Stratify the metric vectors of all customers, and use the clustering method to solve the grouping, with the mathematical expression:

[0117]

[0118] Among them, V i represents the value metric vector of the i-th customer, μ k(i) represents the clustering center to which the i-th customer belongs, N represents the total number of customers, and |·| 2 represents the squared Euclidean distance;

[0119] Specifically, this step fuses the historical customer value data with the current customer behavior information and early warning signals, and uses weighted updating to generate a new customer value metric vector. During the updating process, the smoothing coefficient adjusts the ratio of historical data to newly input data, and the early warning signals intervene with appropriate weights to reflect risk information; then, the clustering method is used to group the metric vectors of each customer, and the Euclidean distance is used to measure the similarity between customers, so as to achieve hierarchical management.

[0120] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating customer value on a cross-border e-commerce platform based on artificial intelligence, characterized in that: including Step S1: Collect order, evaluation, return, and channel information, and perform standardization processing on the collected data to obtain preprocessed data; Step S2: Use natural language processing, image recognition, and speech analysis technologies to process the preprocessed data to extract multimodal features, and use a deep learning model to generate dynamic customer labels; Step S3: Perform time series analysis on the dynamic customer labels to obtain customer behavior time series data; Step S4: Use digital twin technology to perform virtual mapping on the customer behavior time series data and multimodal features to form a customer digital twin image; Step S5: Input the customer behavior time series data into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate warning signals; Step S6: Update customer value evaluation indicators based on the customer behavior time series data and warning signals, and stratify customers.

2. The cross-border e-commerce platform customer value evaluation method based on artificial intelligence according to claim 1, characterized in that: In Step S2, the steps of using natural language processing, image recognition, and speech analysis technologies to process the preprocessed data to extract multimodal features and using a deep learning model to generate dynamic customer labels are as follows: Based on the preprocessed data, embedding methods are used for text, images, and speech respectively to calculate multimodal feature vectors. The defined formula is: F = σ(W1e1 + W2e2 + W3e3 + b), where F represents the multimodal feature vector, e1 represents the text embedding vector, e2 represents the image embedding vector, e3 represents the speech embedding vector, W1 represents the weight matrix corresponding to text embedding, W2 represents the weight matrix corresponding to image embedding, W3 represents the weight matrix corresponding to speech embedding, b represents the bias vector, and σ represents the activation function; Use a fully connected layer to map the multimodal features to the label space, defined as: z = UF + c, L = softmax(z), where z represents the output vector of the fully connected layer, U represents the mapping weight matrix, F represents the aforementioned multimodal feature vector, c represents the mapping bias vector, L represents the dynamic customer label vector, and softmax represents the normalization function.

3. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 2, wherein: The step of extracting the multimodal features also includes processing social interaction data to generate dynamic customer labels containing geographical location, culture, and payment habit information.

4. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 3, wherein: The digital twin image is used to reflect the correspondence between customer behavior time series data and multimodal features.

5. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 4, characterized in that: In Step S4, the steps of using digital twin technology to perform virtual mapping on the customer behavior time series data and multimodal features to form a customer digital twin image are as follows: Aggregate the time-series data of customer behavior {t1, t2, …, t n} by taking the mean. The calculation formula is as follows: Among them, T′ represents the aggregated time-series data vector, and t i represents the feature vector at the i-th moment, and n represents the number of moments; Fuse the aggregated time series data vector T′ and the multimodal feature vector F, and construct a customer digital twin image through a non-linear mapping function. The formula is: D = φ(W t T′ + W f F + d), Among them, D represents the customer digital twin image, W t represents the time-series data conversion weight matrix, T′ represents the aggregated time-series data vector, W f represents the multi-modal feature conversion weight matrix, F represents the multi-modal feature vector, d represents the bias vector, and φ represents the non-linear mapping function.

6. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 5, wherein: In the anomaly detection step of Step S5, an isolation forest algorithm and a deep autoencoder are combined for detection.

7. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 6, wherein: In Step S5, the steps of inputting the customer behavior time series data into a deep autoencoder and an anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate warning signals are as follows: For the customer behavior time series data x, use a deep autoencoder for encoding. The encoding process is defined as: h = f(x) = σ(W e x + b e ), Among them, h represents the encoded feature vector, f(x) represents the encoding function, x represents the input customer behavior time series data vector, W e represents the encoding weight matrix, b e represents the encoding bias vector, and σ represents the activation function; The decoding process reconstructs the encoded vector h into an output vector. The formula is: Among them, represents the reconstructed vector, g(h) represents the decoding function, and W d represents the decoding weight matrix, b d represents the decoding bias vector, and σ represents the activation function.

8. The method for evaluating the customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 7, wherein: In step S5, the step of inputting the customer behavior time series data into the deep autoencoder and the anomaly detection algorithm to detect abnormal return and fraud risk behaviors and generate warning signals further includes: The reconstruction error is calculated using the Euclidean distance, and the formula is: where E represents the reconstruction error, and ‖·‖2 represents the Euclidean distance; Mapping the reconstruction error E to an anomaly score, the formula is: s = ψ(E), where s represents the anomaly score, and ψ represents the score mapping function; After setting the threshold τ, the anomaly score is converted into a warning signal using the indicator function, and the formula is P = 1 {s>τ} , Among them, P represents the warning signal, 1 {s>τ} means taking 1 when s is greater than the threshold τ, otherwise taking 0, and τ represents the detection threshold.

9. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 8, wherein: In step S6, the step of updating the customer value evaluation indicators according to the customer behavior time series data and the warning signals and stratifying the customers is Combine the historical customer value metric V0 with the current behavioral time-series data aggregation vector T agg and the warning signal P, and use a weighted update formula to calculate the updated customer value metric vector V: V = ηV0 + (1 - η)(T agg + δP), Among them, V represents the updated customer value metric vector, V0 represents the previous customer value metric vector, η represents the smoothing coefficient, T agg represents the aggregated vector of behavioral time-series data, δ represents the early warning signal weight, and P represents the early warning signal vector.

10. The method for evaluating customer value of a cross-border e-commerce platform based on artificial intelligence according to claim 9, wherein: In step S6, the step of updating the customer value evaluation indicators according to the customer behavior time series data and the warning signals and stratifying the customers further includes: Stratifying the index vectors of all customers, and using the clustering method to solve the grouping, the mathematical expression is: Among them, V i represents the value index vector of the i-th customer, μ k(i) represents the cluster center to which the i-th customer belongs, N represents the total number of customers, and |·| 2 represents the squared Euclidean distance.

Images