An e-commerce order processing method and system based on artificial intelligence

Abstract

The present invention discloses an e-commerce order processing method and system based on artificial intelligence, and relates to the technical field of e-commerce order processing. It includes collecting order data from an e-commerce platform and preprocessing the collected order data; extracting features from the order data, and using big data processing technology to fuse features to obtain comprehensive feature data; building a multi-layer detection model to detect anomalies on the extracted features, and based on the anomaly detection results, using RPA technology and smart contracts to automatically process abnormal orders; the present invention uses big data processing technology and machine learning algorithms to preprocess, fuse features, detect anomalies, and conduct in-depth analysis on order data, thereby improving the efficiency and accuracy of order processing, and then through the application of natural language processing technology, the system can automatically identify and classify user problems, which not only improves the automation level of order processing, but also optimizes inventory management and marketing strategies.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce order processing, and in particular to an e-commerce order processing method and system based on artificial intelligence. Background Art

[0002] In the daily operation of e-commerce platforms, the complexity of order management continues to increase. This not only involves the receipt, processing and delivery of orders, but also includes the identification and handling of abnormal orders, as well as rapid response and decision support for user issues. Existing order management systems often have difficulty meeting these requirements. They are inefficient in processing large numbers of orders, are not accurate enough in identifying abnormal situations, and have limitations in providing decision support. Summary of the invention

[0003] The purpose of the present invention is to provide an e-commerce order processing method and system based on artificial intelligence, which realizes efficient and intelligent management of e-commerce orders by integrating big data processing, machine learning and natural language processing technologies.

[0004] The purpose of the present invention can be achieved through the following technical solutions:

[0005] This application provides an e-commerce order processing method based on artificial intelligence, comprising the following steps:

[0006] Collect order data from e-commerce platforms and pre-process the collected order data;

[0007] Extract features from order data, use big data processing technology to collect pre-processed data from multiple channels, and perform feature fusion to obtain comprehensive feature data;

[0008] Build a multi-layer detection model to detect anomalies on the extracted features, improve detection accuracy by integrating multiple detection results, and use RPA technology and smart contracts to automatically process abnormal orders based on the anomaly detection results;

[0009] The construction of a multi-layer detection model to perform anomaly detection on the extracted features specifically includes:

[0010] Construct an autoencoder neural network, including an encoder and a decoder, where the encoder is used to compress the input data into a low-dimensional representation, and the decoder is used to restore the low-dimensional representation to the original data; use normal order data to train the autoencoder and adjust the network parameters by minimizing the difference between the input and the reconstructed output;

[0011] Use the trained autoencoder to reconstruct the new order data and calculate the reconstruction error using the Huber loss function;

[0012] The threshold is set to distinguish normal data from abnormal data. When the reconstruction error exceeds the threshold, the data point is considered abnormal.

[0013] Based on the reconstruction error and the set threshold, abnormal orders are identified and then processed automatically using RPA technology and smart contracts;

[0014] Use natural language processing technology to identify and classify questions raised by users related to abnormal orders, match corresponding solutions, and then generate natural and fluent reply content based on the classification results.

[0015] Furthermore, before collecting order data, the legitimacy of the information acquisition request is verified through an authentication mechanism, including:

[0016] The requester sends an authentication request to the authentication server through the OpenID Connect protocol. The authentication server verifies the identity credentials of the requester. When the identity authentication is successful, the authentication server returns an ID token containing the identity information of the requester.

[0017] Data access permissions are defined based on predefined roles or attributes. The requester's identity token is used to determine its role and attributes and compared with the permission list. When the requester's role or attribute matches the access permissions for the required data, data access permissions are granted.

[0018] Furthermore, big data processing technology is used to collect pre-processed data from multiple channels and perform feature fusion to obtain comprehensive feature data, including:

[0019] Use the chi-square test to evaluate the correlation between each feature and the target variable, and select features with high correlation with the target variable;

[0020] Input the features with high correlation with the target variable into the feature fusion algorithm of sparse representation theory to generate comprehensive feature data;

[0021] Among them, the features selected from the chi-square test method are used as input data to construct a dictionary matrix, in which each column represents a feature, initialize the sparse representation coefficient vector, and use the Lasso regression algorithm to optimize the sparse representation coefficient vector, which is expressed as: ,in is a vector of raw feature data, containing features selected from the chi-square test that have a high correlation with the target variable. represents the dictionary matrix, is the feature matrix, is the sparse representation coefficient vector, is a regularization parameter that controls the strength of the L1 regularization term, thereby promoting sparsity, represents the L1 norm of the coefficient vector, that is, the sum of the absolute values ​​of all coefficients,

[0022] After using the Lasso regression algorithm to optimize the sparse representation coefficient vector, the comprehensive feature data is expressed as: .

[0023] Furthermore, the reconstruction error is calculated by the Huber loss function, which is specifically expressed as:

[0024] ,in is the original input data, is the reconstructed output of the autoencoder, is a hyperparameter that controls the sensitivity of the loss function when the error is large.

[0025] Furthermore, RPA technology and smart contracts are used to automatically process abnormal orders, including:

[0026] After identifying abnormal orders, use RPA technology and smart contracts to automatically process abnormal orders and pass the detected abnormal order information to the automation system;

[0027] RPA technology is used to simulate human user operations, automatically log in to the e-commerce platform, access the order management system, and execute specific processing procedures based on the characteristics of abnormal orders. At the same time, smart contracts automatically execute preset conditions and rules on the blockchain. When an order is marked as abnormal, the abnormal order is automatically triggered and executed.

[0028] Furthermore, natural language processing technology is used to identify and classify questions raised by users related to abnormal orders and match corresponding solutions, including:

[0029] The SimCSE model calculates the similarity of texts, matches user questions with predefined question categories, and identifies and classifies questions;

[0030] The SimCSE model is expressed as:

[0031] ,in Represents text The non-negative semantic distance between is the adjustment factor, and the SimCSE model is meaningful when the semantic distance is 0;

[0032] Based on the classification results, the corresponding solutions are retrieved from the knowledge base, and the solutions are constructed into natural language replies using natural language generation technology, and then sent to the user through an automated process.

[0033] Furthermore, it also includes: using machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, and automatically generate the optimal order processing strategy based on the analysis and modeling results.

[0034] Furthermore, we use machine learning algorithms and data mining techniques to conduct in-depth analysis of order data to identify purchasing patterns and user behaviors, including:

[0035] Use regression analysis to predict order amounts and prioritize based on user classification;

[0036] Use time series forecasting algorithms to predict future demand for products and adjust inventory based on user purchasing patterns;

[0037] Based on the user's purchase history and behavior patterns, matrix decomposition is used to automatically generate personalized recommendation strategies.

[0038] Furthermore, it also includes: combining big data and artificial intelligence technologies, continuously monitoring and optimizing the processing process, dynamically adjusting strategies, and forming a closed-loop feedback mechanism.

[0039] The present application also provides an e-commerce order processing system based on artificial intelligence, which is used to implement an e-commerce order processing method based on artificial intelligence, specifically including:

[0040] The data preprocessing module collects order data from the e-commerce platform and cleans, normalizes, processes missing values, and analyzes outliers on the collected order data. The identity of the requester is verified through the OpenID Connect protocol before data collection.

[0041] The feature fusion module uses big data processing technology, combined with the chi-square test and Lasso regression algorithm, to collect pre-processed data from multiple channels, perform feature fusion, and generate comprehensive feature data that reflects key information of the order;

[0042] The anomaly detection module builds an autoencoder neural network, uses the Huber loss function to calculate the reconstruction error, identifies abnormal orders, and automatically processes them through RPA technology and smart contracts;

[0043] The natural language processing module uses the SimCSE model to calculate text similarity, classify user questions, retrieve solutions from the knowledge base, and automatically generate natural and fluent reply content;

[0044] The analysis and decision support module uses machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, predict order amounts and product demand, and automatically generate the optimal order processing strategy.

[0045] The beneficial effects of the present invention are:

[0046] The present invention uses advanced big data processing technology to automatically collect order data from multiple channels, and performs cleaning, normalization, processing of missing values ​​and outliers, as well as data deduplication and anomaly detection. It significantly improves the efficiency of order processing and solves the problem that traditional order management systems often require a lot of manual operations to process and identify abnormal orders, which is not only time-consuming but also prone to errors. Through feature extraction and fusion technology, it can automatically select features with high correlation with the target variable, and through the sparse representation characteristics of Lasso regression, generate a new feature set that retains key information and has a lower dimension. This method not only improves the prediction performance and computational efficiency of the model, but also reduces the resources required for model training, further improving the efficiency of order processing;

[0047] By building a multi-layer detection model, including an autoencoder neural network and an anomaly detection mechanism based on the Huber loss function, the accuracy of identifying abnormal orders is effectively improved. The autoencoder neural network learns the internal representation of normal order data through training, and uses the Huber loss function to calculate the reconstruction error, thereby distinguishing normal and abnormal data. Once an anomaly is detected, RPA technology and smart contracts automatically intervene to quickly process these orders, ensuring efficient problem solving and response, which not only improves the accuracy of anomaly detection, but also speeds up processing and reduces the need for manual intervention. In this way, e-commerce platforms can more accurately identify and process abnormal orders, reducing customer dissatisfaction and potential economic losses caused by incorrect processing;

[0048] By using natural language processing technology, especially the SimCSE model to calculate text similarity, the e-commerce platform can accurately match the questions raised by users related to abnormal orders with predefined question categories, thereby identifying and classifying the questions, and then retrieve the corresponding solutions from the knowledge base based on the calculated similarity, and use natural language generation technology to build the solutions into natural language replies, which are then sent to users through an automated process. This not only improves the efficiency of abnormal order processing, but also enhances the user service experience. Users can quickly obtain clear and accurate feedback, which improves satisfaction and loyalty. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] For better understanding and implementation, the technical solution of the present application is described in detail below with reference to the accompanying drawings.

[0050] Figure 1 A flowchart of an artificial intelligence-based e-commerce order processing method provided in Example 1 of the present application;

[0051] Figure 2A schematic diagram of a process for performing feature fusion to obtain comprehensive feature data in an e-commerce order processing method based on artificial intelligence provided in Example 1 of the present application;

[0052] Figure 3 A schematic diagram of a process for constructing a multi-layer detection model to perform anomaly detection on extracted features in an e-commerce order processing method based on artificial intelligence provided in Example 1 of the present application;

[0053] Figure 4 A schematic diagram of the structure of an artificial intelligence-based e-commerce order processing system provided in Example 2 of the present application. DETAILED DESCRIPTION

[0054] In order to further explain the technical means and effects taken by the present invention to achieve the predetermined invention purpose, exemplary embodiments will be described in detail here, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are only examples of methods and systems consistent with some aspects of the present application as detailed in the attached claims.

[0055] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0056] The specific implementation methods, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0057] Example 1

[0058] See also Figure 1-Figure 3 This embodiment provides an e-commerce order processing method and system based on artificial intelligence, which realizes efficient and intelligent management of e-commerce orders by integrating big data processing, machine learning and natural language processing technologies.

[0059] The present invention provides an e-commerce order processing method based on artificial intelligence, comprising the following steps:

[0060] S1. Collect order data from the e-commerce platform and pre-process the collected order data;

[0061] Furthermore, before collecting order data, the legitimacy of the information acquisition request is verified through an authentication mechanism, including:

[0062] The requester sends an authentication request to the authentication server through the OpenID Connect protocol. The authentication server verifies the identity credentials of the requester, such as username and password. When the identity authentication is successful, the authentication server returns an ID token containing the identity information of the requester.

[0063] Data access permissions are defined based on predefined roles or attributes, such as user roles, departments, or data access levels. The requester's identity token is used to determine its role and attributes and compare them with the permission list. When the requester's role or attributes match the access permissions for the required data, data access permissions are granted.

[0064] Specifically, by implementing the OpenID Connect protocol for identity authentication, the e-commerce platform can ensure that only verified requesters can access order data, thereby enhancing data security. At the same time, by defining data access permissions based on predefined roles or attributes, the system can accurately control the access levels of different users to specific data. When the requester's identity token matches its role and attributes, the corresponding data access permissions are granted, which ensures both the accessibility of the data and the compliance and security of data access.

[0065] The preprocessing of the collected order data includes: cleaning the order data, removing invalid and redundant records, and then normalizing the order data, including unifying the field format, converting data types, such as converting timestamps to standard date formats, and processing missing values. In addition, outliers are analyzed, such as abnormal fluctuations in order amounts, and processed accordingly; finally, data deduplication and anomaly detection are performed to ensure the accuracy of subsequent analysis and model building. The preprocessing process of order data provides the e-commerce platform with high-quality, standardized data to support subsequent data analysis and decision-making.

[0066] S2. Extract features from order data, use big data processing technology to collect pre-processed data from multiple channels, and perform feature fusion to obtain comprehensive feature data;

[0067] Furthermore, big data processing technology is used to collect pre-processed data from multiple channels and perform feature fusion to obtain comprehensive feature data, including:

[0068] Use the chi-square test to evaluate the correlation between each feature and the target variable, and select features with high correlation with the target variable;

[0069] Input the features with high correlation with the target variable into the feature fusion algorithm of sparse representation theory to generate comprehensive feature data;

[0070] Among them, the features selected from the chi-square test method are used as input data to construct a dictionary matrix, in which each column represents a feature, initialize the sparse representation coefficient vector, and use the Lasso regression algorithm to optimize the sparse representation coefficient vector, which is expressed as: ,in is a vector of raw feature data, containing features selected from the chi-square test that have a high correlation with the target variable. represents the dictionary matrix, is the feature matrix, is the sparse representation coefficient vector, is a regularization parameter that controls the strength of the L1 regularization term, thereby promoting sparsity, represents the L1 norm of the coefficient vector, that is, the sum of the absolute values ​​of all coefficients,

[0071] After using the Lasso regression algorithm to optimize the sparse representation coefficient vector, the comprehensive feature data is expressed as: .

[0072] By minimizing the above objective function, the Lasso regression algorithm not only tries to find the coefficients that can best reconstruct the original feature data (that is, minimize the residual sum of squares), but also makes the coefficient vector sparse through the L1 regularization term, that is, making as many coefficients as possible zero. In this way, the final coefficient vector It can not only effectively represent the original feature data, but also automatically perform feature selection to retain only the most important features.

[0073] Specifically, the effect of combining the chi-square test and the Lasso regression algorithm for feature fusion is that it can effectively identify and select features that are highly correlated with the target variable, and through the sparse representation characteristics of the Lasso regression, generate a new feature set that retains key information and has a lower dimension. The advantage of this method is that it not only uses the statistical significance of the chi-square test to screen features, but also promotes the generalization ability of the model and prevents overfitting through the L1 regularization of the Lasso regression, while achieving feature selection and the generation of comprehensive feature data. The result of this is to improve the predictive performance and computational efficiency of the model, while reducing the resources required for model training.

[0074] S3. Build a multi-layer detection model to detect anomalies on the extracted features, improve detection accuracy by integrating multiple detection results, and use RPA technology and smart contracts to automatically process abnormal orders based on the anomaly detection results to ensure rapid response and resolution;

[0075] Furthermore, a multi-layer detection model is constructed to perform anomaly detection on the extracted features, specifically including:

[0076] S31. Construct an autoencoder neural network, including an encoder and a decoder, wherein the encoder is used to compress input data into a low-dimensional representation, and the decoder is used to restore the low-dimensional representation to the original data; use normal order data to train the autoencoder, and adjust the network parameters by minimizing the difference between the input and the reconstructed output. After the training is completed, the autoencoder can learn to accurately reconstruct the internal representation of normal data;

[0077] S32, reconstruct the new order data using the trained autoencoder, and calculate the reconstruction error, i.e., the difference between the original input data and the reconstructed output of the autoencoder, using the Huber loss function;

[0078] The threshold is set to distinguish normal data from abnormal data. When the reconstruction error exceeds the threshold, the data point is considered abnormal.

[0079] S33. Identify abnormal orders based on reconstruction errors and set thresholds, and then use RPA technology and smart contracts to automatically process abnormal orders.

[0080] Specifically, by building a multi-layer detection model, including an autoencoder neural network and an anomaly detection mechanism based on the Huber loss function, the e-commerce platform can effectively identify abnormal orders. The autoencoder learns the internal representation of normal orders by compressing and reconstructing order data, while the Huber loss function is used to calculate the reconstruction error to distinguish between normal and abnormal data. Once an anomaly is detected, RPA technology and smart contracts automatically intervene to quickly process these orders, ensuring efficient problem solving and response. This approach not only improves the accuracy of anomaly detection, but also speeds up processing and reduces the need for manual intervention.

[0081] Furthermore, the reconstruction error is calculated by the Huber loss function, which is specifically expressed as:

[0082] ,in is the original input data, is the reconstructed output of the autoencoder, is a hyperparameter that controls the sensitivity of the loss function when the error is large.

[0083] Using the Huber loss function to calculate the reconstruction error can effectively balance the bias and variance in anomaly detection. By adjusting the hyperparameters to adapt to different degrees of error, it can remain sensitive to small errors while being robust to large errors. This helps the autoencoder to more accurately identify normal and abnormal order data.

[0084] Furthermore, RPA technology and smart contracts are used to automatically process abnormal orders, including:

[0085] After identifying abnormal orders, use Robotic Process Automation (RPA) technology and smart contracts to automatically process abnormal orders and pass the detected abnormal order information to the automated system;

[0086] RPA technology is used to simulate human user operations, automatically log in to the e-commerce platform, access the order management system, and execute specific processing procedures based on the characteristics of abnormal orders, such as refunds, order cancellations, or contacting customers. At the same time, smart contracts automatically execute preset conditions and rules on the blockchain to ensure consistency and transparency of processing. When an order is marked as abnormal, it is automatically triggered and executed, thereby quickly responding to and resolving abnormal situations, reducing manual intervention, and improving processing efficiency and accuracy.

[0087] Specifically, by combining robotic process automation (RPA) technology and smart contracts, e-commerce platforms can achieve efficient and automated processing of abnormal orders. When the autoencoder neural network identifies an abnormal order, RPA technology automatically performs a series of operations, such as logging into the e-commerce platform, accessing the order management system, and executing preset processes such as refunds, order cancellations, or contacting customers based on the characteristics of the abnormal order. At the same time, smart contracts automatically execute relevant conditions and rules on the blockchain to ensure consistency and transparency of the processing process. This automated process not only greatly reduces manual intervention, but also improves processing efficiency and accuracy, ensuring rapid response and resolution of abnormal situations.

[0088] S4. Use natural language processing technology to identify and classify questions raised by users related to abnormal orders, match corresponding solutions, and then generate natural and fluent reply content based on the classification results to improve user service experience.

[0089] Furthermore, natural language processing technology is used to identify and classify questions raised by users related to abnormal orders and match corresponding solutions, including:

[0090] The SimCSE model calculates the similarity of texts, matches user questions with predefined question categories, and identifies and classifies questions;

[0091] The SimCSE model is expressed as:

[0092] ,in Represents text The non-negative semantic distance between is the adjustment factor, and the SimCSE model is meaningful when the semantic distance is 0;

[0093] Based on the classification results, the corresponding solutions are retrieved from the knowledge base, and the solutions are constructed into natural language replies using natural language generation technology, and then sent to the user through an automated process.

[0094] Specifically, by using natural language processing technology, especially the SimCSE model to calculate text similarity, the e-commerce platform can accurately match the questions raised by users related to abnormal orders with predefined question categories, thereby identifying and classifying the questions. First, it involves splitting the text into words and removing stop words, then mapping each word to an independent dimension, and calculating the word frequency of each dimension to form a word frequency matrix; then using the contrastive learning principle in the SimCSE model, by narrowing the distance between similar data and widening the distance between dissimilar data, it can better learn the representation of data and produce better results in text matching tasks. Finally, the system will retrieve the corresponding solution from the knowledge base based on the calculated similarity, and use natural language generation technology to build the solution into a natural language reply, and then send it to the user through an automated process to ensure that they quickly get clear and accurate feedback. This process not only improves the efficiency of abnormal order processing, but also significantly improves the user service experience, ensuring that users can quickly get accurate and useful feedback.

[0095] S5. Use machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, and automatically generate the optimal order processing strategy based on the analysis and modeling results to provide decision support for enterprises.

[0096] Furthermore, we use machine learning algorithms and data mining techniques to conduct in-depth analysis of order data to identify purchasing patterns and user behaviors, including:

[0097] Use regression analysis to predict order amounts and prioritize based on user classification;

[0098] Use time series forecasting algorithms to predict future demand for products and adjust inventory based on user purchasing patterns;

[0099] Based on the user's purchase history and behavior patterns, matrix decomposition is used to automatically generate personalized recommendation strategies.

[0100] Specifically, by using machine learning algorithms and data mining techniques, companies can deeply analyze order data to identify users' purchasing patterns and behavioral characteristics. Based on the collection of data such as users' purchase records, browsing history, and interactive behaviors, data preprocessing is then used to ensure its quality and consistency. Next, regression analysis is used to predict the order amount, and user classification is combined to prioritize to determine which orders contribute the most to revenue. Then, a time series forecasting algorithm is used to predict future demand for products, and inventory is adjusted in combination with user purchasing patterns to ensure that inventory levels match market demand. Finally, based on users' purchasing history and behavioral patterns, matrix decomposition technology is used to automatically generate personalized recommendation strategies to improve user satisfaction and purchase conversion rates, which together constitute a complete analysis process that enables companies to automatically generate optimal order processing strategies based on data-driven insights, optimize inventory management, personalized recommendations, and improve marketing effectiveness.

[0101] S6. Combine big data and artificial intelligence technologies to continuously monitor and optimize the processing process, dynamically adjust strategies, and form a closed-loop feedback mechanism.

[0102] This embodiment can effectively improve the intelligent level of order processing through the e-commerce platform. First, the security of data access is ensured through the OpenIDConnect protocol, and data access rights are defined based on predefined roles or attributes to ensure that only authorized users can access sensitive data. Then, the collected order data is preprocessed, including cleaning, normalization, processing of missing values ​​and outliers, as well as data deduplication and anomaly detection, to ensure data quality and consistency, and provide a basis for subsequent data analysis and model building;

[0103] Using big data processing technology, pre-processed data is collected from multiple channels and feature fusion is performed to generate comprehensive feature data. Through the chi-square test and Lasso regression algorithm, features with high correlation with the target variable are effectively identified and selected to generate a new feature set that retains key information and has a lower dimension. This not only improves the model's predictive performance and computational efficiency, but also reduces the resources required for model training;

[0104] A multi-layer detection model is built, including an autoencoder neural network, to detect anomalies on the extracted features. The autoencoder learns the internal representation of normal orders by compressing and reconstructing order data, while the Huber loss function is used to calculate the reconstruction error to distinguish between normal and abnormal data. Once an anomaly is detected, RPA technology and smart contracts automatically intervene to quickly process these orders, ensuring efficient problem resolution and response;

[0105] Use natural language processing technology, especially the SimCSE model, to calculate text similarity and accurately match users' questions related to abnormal orders with predefined question categories to identify and classify questions. The system will retrieve the corresponding solutions from the knowledge base based on the calculated similarity, and use natural language generation technology to build the solutions into natural language replies, which will then be sent to users through an automated process to improve user service experience;

[0106] Use machine learning algorithms and data mining techniques to conduct in-depth analysis of order data and identify purchasing patterns and user behaviors. Through regression analysis, time series prediction, and matrix decomposition techniques, automatically generate optimal order processing strategies to provide decision support for enterprises. These strategies help optimize inventory management, personalize recommendations, and improve marketing effectiveness;

[0107] Finally, by combining big data and artificial intelligence technologies, continuously monitoring and optimizing the processing process, dynamically adjusting strategies, and forming a closed-loop feedback mechanism, not only does it improve the efficiency of abnormal order processing, it also significantly improves the user service experience, ensuring that users can quickly obtain accurate and useful feedback, thereby enhancing the company's market competitiveness and operational efficiency.

[0108] Example 2

[0109] refer to Figure 4 This embodiment provides an e-commerce order processing system based on artificial intelligence, which is used to implement an e-commerce order processing method based on artificial intelligence, specifically including:

[0110] The data preprocessing module collects order data from the e-commerce platform and cleans, normalizes, processes missing values, and analyzes outliers on the collected order data. The identity of the requester is verified through the OpenID Connect protocol before data collection.

[0111] The feature fusion module uses big data processing technology, combined with the chi-square test and Lasso regression algorithm, to collect pre-processed data from multiple channels, perform feature fusion, and generate comprehensive feature data that reflects key information of the order;

[0112] The anomaly detection module builds an autoencoder neural network, uses the Huber loss function to calculate the reconstruction error, identifies abnormal orders, and automatically processes them through RPA technology and smart contracts;

[0113] The natural language processing module uses the SimCSE model to calculate text similarity, classify user questions, retrieve solutions from the knowledge base, and automatically generate natural and fluent reply content;

[0114] The analysis and decision support module uses machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, predict order amounts and product demand, and automatically generate the optimal order processing strategy.

[0115] The artificial intelligence-based e-commerce order processing system in this embodiment realizes comprehensive intelligent management of order data through integrated data preprocessing, feature fusion, anomaly detection, natural language processing, and analysis and decision support modules. It not only improves the efficiency and accuracy of order processing, but also reduces manual intervention through automation technology, optimizes user experience, and enhances the market competitiveness of enterprises through data-driven decision support.

[0116] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An e-commerce order processing method based on artificial intelligence, characterized by: The steps include: Collect order data from e-commerce platforms and pre-process the collected order data; Extract features from order data, use big data processing technology to collect pre-processed data from multiple channels, and perform feature fusion to obtain comprehensive feature data; Build a multi-layer detection model to detect anomalies on the extracted features. Based on the anomaly detection results, use RPA technology and smart contracts to automatically process abnormal orders. The construction of a multi-layer detection model to perform anomaly detection on the extracted features specifically includes: Construct an autoencoder neural network, including an encoder and a decoder, where the encoder is used to compress the input data into a low-dimensional representation, and the decoder is used to restore the low-dimensional representation to the original data; use normal order data to train the autoencoder and adjust the network parameters by minimizing the difference between the input and the reconstructed output; Use the trained autoencoder to reconstruct the new order data and calculate the reconstruction error using the Huber loss function; The threshold is set to distinguish normal data from abnormal data. When the reconstruction error exceeds the threshold, the data point is considered abnormal. Based on the reconstruction error and the set threshold, abnormal orders are identified and then processed automatically using RPA technology and smart contracts; Use natural language processing technology to identify and classify questions raised by users related to abnormal orders, match corresponding solutions, and then generate natural and fluent replies based on the classification results; Among them, big data processing technology is used to collect pre-processed data from multiple channels and perform feature fusion to obtain comprehensive feature data, including: Use the chi-square test to evaluate the correlation between each feature and the target variable, and select features with high correlation with the target variable; Input the features with high correlation with the target variable into the feature fusion algorithm of sparse representation theory to generate comprehensive feature data; The features selected from the chi-square test method are used as input data to construct a dictionary matrix, where each column represents a feature, initialize the sparse representation coefficient vector, and use the Lasso regression algorithm to optimize the sparse representation coefficient vector, which is expressed as: ,in is a vector of raw feature data, containing features selected from the chi-square test that have a high correlation with the target variable. represents the dictionary matrix, is the feature matrix, is the sparse representation coefficient vector, is a regularization parameter that controls the strength of the L1 regularization term, thereby promoting sparsity, represents the L1 norm of the coefficient vector, that is, the sum of the absolute values ​​of all coefficients, After using the Lasso regression algorithm to optimize the sparse representation coefficient vector, the comprehensive feature data is expressed as: ; Among them, the reconstruction error is calculated by the Huber loss function, which is expressed as: , where is the original input data, is the reconstructed output of the autoencoder, is a hyperparameter used to control the sensitivity of the loss function to errors; Among them, natural language processing technology is used to identify and classify questions related to abnormal orders raised by users and match corresponding solutions, including: The SimCSE model calculates the similarity of texts, matches user questions with predefined question categories, and identifies and classifies questions; The SimCSE model is expressed as: , where Represents text The non-negative semantic distance between is the adjustment factor, and the SimCSE model is meaningful when the semantic distance is 0; Based on the classification results, the corresponding solutions are retrieved from the knowledge base, and the solutions are constructed into natural language replies using natural language generation technology, and then sent to the user through an automated process.

2. The method for processing e-commerce orders based on artificial intelligence according to claim 1, characterized in that: Before collecting order data, the legitimacy of the information acquisition request is verified through an authentication mechanism, including: The requester sends an authentication request to the authentication server through the OpenID Connect protocol. The authentication server verifies the identity credentials of the requester. When the identity authentication is successful, the authentication server returns an ID token containing the identity information of the requester. Data access permissions are defined based on predefined roles or attributes. The requester's identity token is used to determine its role and attributes and compared with the permission list. When the requester's role or attribute matches the access permissions for the required data, data access permissions are granted.

3. The method for processing e-commerce orders based on artificial intelligence according to claim 1, characterized in that: Use RPA technology and smart contracts to automatically process abnormal orders, including: After identifying abnormal orders, use RPA technology and smart contracts to automatically process abnormal orders and pass the detected abnormal order information to the automation system; RPA technology is used to simulate human user operations, automatically log in to the e-commerce platform, access the order management system, and execute specific processing procedures based on the characteristics of abnormal orders. At the same time, smart contracts automatically execute preset conditions and rules on the blockchain. When an order is marked as abnormal, the abnormal order is automatically triggered and executed.

4. The method for processing e-commerce orders based on artificial intelligence according to claim 1, characterized in that: Also includes: Use machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, and automatically generate the optimal order processing strategy based on the analysis and modeling results.

5. The method for processing e-commerce orders based on artificial intelligence according to claim 4, characterized in that: Using machine learning algorithms and data mining techniques, we conduct in-depth analysis of order data to identify purchasing patterns and user behaviors, including: Use regression analysis to predict order amounts and prioritize based on user classification; Use time series forecasting algorithms to predict future demand for products and adjust inventory based on user purchasing patterns; Based on the user's purchase history and behavior patterns, matrix decomposition is used to automatically generate personalized recommendation strategies.

6. The method for processing e-commerce orders based on artificial intelligence according to claim 1, characterized in that: Also includes: Combining big data and artificial intelligence technologies, the processing process is continuously monitored and optimized, and strategies are dynamically adjusted to form a closed-loop feedback mechanism.

7. An artificial intelligence-based e-commerce order processing system, used to implement an artificial intelligence-based e-commerce order processing method as described in any one of claims 1 to 6, characterized in that: include: The data preprocessing module collects order data from the e-commerce platform and cleans, normalizes, processes missing values, and analyzes outliers on the collected order data. The identity of the requester is verified through the OpenID Connect protocol before data collection. The feature fusion module uses big data processing technology, combined with the chi-square test and Lasso regression algorithm, to collect pre-processed data from multiple channels, perform feature fusion, and generate comprehensive feature data that reflects key information of the order; The anomaly detection module builds an autoencoder neural network, uses the Huber loss function to calculate the reconstruction error, identifies abnormal orders, and automatically processes them through RPA technology and smart contracts; The natural language processing module uses the SimCSE model to calculate text similarity, classify user questions, retrieve solutions from the knowledge base, and automatically generate natural and fluent reply content; The analysis and decision support module uses machine learning algorithms and data mining technology to conduct in-depth analysis of order data, identify purchasing patterns and user behaviors, predict order amounts and product demand, and automatically generate the optimal order processing strategy.

Images