SVM-based wild animal and plant illegal transaction e-commerce platform supervision risk data mining method

Abstract

The invention relates to the technical field of transaction e-commerce platform supervision, and discloses a wild animal and plant illegal transaction e-commerce platform supervision risk data mining method based on an SVM, and the method comprises the following steps: data collection, data preprocessing, feature selection, SVM model training, risk assessment, supervision decision support, model updating and data security protection. According to the wild animal and plant illegal transaction e-commerce platform supervision risk data mining method based on the SVM, an efficient, accurate and real-time supervision system is constructed through multi-dimensional data collection, preprocessing, feature extraction, model training and risk assessment. The system not only can automatically identify and mark high-risk illegal transactions, but also can help a supervision mechanism to quickly respond and formulate an effective supervision strategy through real-time early warning and a visual tool. In addition, the self-adaptive capability and data security protection measures of the system ensure that the user privacy can be continuously optimized and protected in the face of continuously changing illegal transaction modes.

Description

Technical Field

[0001] The present invention relates to the technical field of supervision of trading e-commerce platforms, and specifically to a method for mining supervision risk data of illegal wildlife trading e-commerce platforms based on SVM. Background Art

[0002] With the rapid development of Internet technology, e-commerce platforms have become an important channel for commodity trading. However, this convenience has also provided new ways for illegal wildlife trading. Due to the anonymity and cross-regional nature of online transactions, regulatory agencies face huge challenges in combating and preventing illegal wildlife trading. Currently, the supervision of most e-commerce platforms relies on manual review, which is inefficient and difficult to cover all transactions, resulting in illegal trading activities being difficult to be discovered and processed in a timely manner.

[0003] Support Vector Machine (SVM) is an effective supervised learning method, widely used in fields such as pattern recognition, classification, and regression analysis. SVM can correctly classify sample points in a dataset by constructing a hyperplane as the decision boundary, especially performing well in dealing with high-dimensional data. However, applying SVM to the mining of supervision risk data for illegal wildlife trading, there are currently no mature technologies and methods. Summary of the Invention

[0004] (I) Technical Problems to be Solved Aiming at the deficiencies of the prior art, the present invention provides a method for mining supervision risk data of illegal wildlife trading e-commerce platforms based on SVM to solve the above problems.

[0005] (II) Technical Solutions To achieve the above object, the present invention provides the following technical solution: A method for mining supervision risk data of illegal wildlife trading e-commerce platforms based on SVM, including the following steps: Data collection: Collect transaction data from e-commerce platforms, where the transaction data includes commodity information, user information, and transaction records; Data preprocessing: Clean the collected data, remove duplicate, invalid or abnormal data, and perform formatting processing, and extract key features through natural language processing (NLP) and image processing technologies; Feature selection: Use methods such as statistical analysis and information gain to select features related to illegal trading risks from the preprocessed data; SVM model training: Use the selected features and known illegal trading cases as the training set to train the SVM model. During the training process, use the kernel function to map the data to a high-dimensional space to find the optimal classification hyperplane, and improve the generalization ability and classification accuracy of the model through cross-validation and parameter tuning; Risk assessment: Apply the trained SVM model to real-time transaction data to conduct risk assessment for each transaction. The model will calculate the probability that the transaction belongs to the illegal transaction category based on the extracted features, and determine the risk level of the transaction according to the set threshold; Regulatory decision support: Provide decision support for regulatory agencies based on the risk assessment results, including marking high-risk transactions, automatic alarm, and suggesting regulatory measures; Model update: Regularly update the SVM model according to new illegal transaction cases and transaction data to improve the accuracy and adaptability of the model; Data security protection: Take measures such as encryption and access control to ensure the security and confidentiality of the transaction data collected and processed.

[0006] Preferably, feature extraction includes: Use NLP technology to extract keywords and phrases from product descriptions and identify words related to wild animals and plants; Use image processing technology to identify species features in product pictures, such as shape, color, texture, etc.; Analyze the user's historical transaction records and extract features of abnormal transaction behaviors, such as frequently trading high-risk products, abnormal fluctuations in transaction amounts, etc.; Consider the user's geographical location information and analyze the transaction risk characteristics in different regions.

[0007] Preferably, SVM model construction includes: Select appropriate kernel functions and parameters to map the training data to a high-dimensional space; Adopt the cross-validation method to evaluate the model performance and select the optimal combination of kernel functions and parameters; Normalize the training data to ensure that each feature has the same weight in model training; Introduce a regularization term to prevent the model from overfitting.

[0008] Preferably, risk assessment and decision support include: Set a reasonable risk threshold and judge the risk level of the transaction according to the probability value output by the model; Develop an alarm system to monitor high-risk transactions in real time and automatically send alarm messages; Provide a visualization tool to display the risk assessment results and transaction trends, and assist regulatory agencies in formulating regulatory strategies; Establish a risk early warning mechanism to predict and warn of potential illegal transaction trends.

[0009] Preferably, feature extraction also uses deep learning technology to perform more accurate species identification and feature extraction on product pictures.

[0010] Preferably, the SVM model construction also adopts an ensemble learning method to combine multiple SVM models, improving the stability and accuracy of the model.

[0011] Preferably, the risk assessment and decision support also customize the risk assessment indicators and thresholds according to different regulatory requirements and scenarios.

[0012] Compared with the prior art, the present invention provides a method for mining regulatory risk data of e-commerce platforms for illegal wildlife trading based on SVM, having the following beneficial effects: The method for mining regulatory risk data of e-commerce platforms for illegal wildlife trading based on SVM constructs an efficient, accurate, and real-time regulatory system through multi-dimensional data collection, preprocessing, feature extraction, model training, and risk assessment. This system can not only automatically identify and mark high-risk illegal transactions, but also, through real-time early warning and visualization tools, help regulatory agencies respond quickly and formulate effective regulatory strategies. In addition, the system's adaptive ability and data security protection measures ensure that it can continuously optimize and protect user privacy when facing constantly changing illegal trading patterns.

[0013] The implementation of this method not only improves the regulatory ability of e-commerce platforms for illegal wildlife trading, but also provides a technical path that can be used for reference in the supervision of illegal transactions in other fields. By combining advanced machine learning technologies and data analysis methods, this system has laid a solid foundation for building a more intelligent, efficient, and secure e-commerce platform supervision system. In the future, with the further development of technology and the continuous accumulation of data, this system is expected to play an important role in more fields and promote the development of e-commerce platform supervision towards a more intelligent and refined direction. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the operation step flow of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The present invention will be further described in detail below in conjunction with the drawings and embodiments. Obviously, the described embodiments are only a part of the embodiments of the invention, rather than all of the embodiments. Without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the invention without creative efforts shall fall within the scope of protection of the invention.

[0016] In addition, "multiple" means more than two. In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or is unable to be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the invention.

[0017] Please refer to Figure 1 , a method for mining regulatory risk data of e-commerce platforms for illegal wildlife trading based on SVM: I. Data collection Cooperate with e-commerce platforms Establish a data sharing agreement with major e-commerce platforms to ensure legal access to transaction data. These e-commerce platforms cover various product categories that may involve wildlife trading, including but not limited to handicrafts, medicinal materials, pets and other related categories.

[0018] Extract transaction data from the database of e-commerce platforms, including product information (such as product title, description, price, category, etc.), user information (such as username, registration address, contact information, etc.) and transaction records (such as transaction time, transaction amount, transaction quantity, etc.).

[0019] II. Data preprocessing Data cleaning Write a data cleaning program to preliminarily screen the collected data. Remove obviously duplicate data records, such as the situation where the same product is recorded multiple times in a short period and all information is exactly the same.

[0020] Identify and remove invalid data, such as data that lacks key descriptions in product information (such as the product name is empty) or data that does not conform to logic such as a negative transaction amount in the transaction record.

[0021] For abnormal data, use statistical methods for detection. For example, when the transaction amount of a certain user far exceeds the normal price range of this type of product (determine the normal range by calculating the mean and standard deviation of the price of this type of product), mark it as abnormal data and conduct further analysis or removal.

[0022] Formatting processing Unify the conversion of data in different formats into a format suitable for subsequent processing. Unify the date format to "YYYY-MM-DD" and unify the number format to a specific precision (such as retaining two decimal places).

[0023] Key feature extraction Use NLP technology to extract product description features Process the product descriptions using natural language processing toolkits such as NLTK or StanfordNLP. First, perform lexical analysis to break down the description text into words and phrases.

[0024] Build a vocabulary related to wildlife. By matching the words in the vocabulary, identify keywords and phrases related to wildlife. For example, "ivory", "rhinoceros horn", "pangolin scales", etc. are words related to illegal wildlife products, and "orchid", "cycad", etc. are words related to protected wild plants.

[0025] Use image processing technology to extract the features of product images For product images, use an image processing library (such as OpenCV) for preliminary processing. Resize the images to a unified size and resolution for subsequent analysis.

[0026] Use image feature extraction algorithms such as SIFT (Scale-Invariant Feature Transform) or HOG (Histogram of Oriented Gradients) to extract features such as shape, color, and texture in the images. At the same time, use deep learning technology (such as pre-trained convolutional neural network models such as ResNet or VGGNet) to perform more accurate species recognition and feature extraction on product images. Input the images into the pre-trained model to obtain the feature vectors output by the model, which can more accurately reflect the species information in the images.

[0027] Analyze the characteristics of the user's historical transaction records From the user's historical transaction records, count information such as transaction frequency and the distribution of transaction amounts. By calculating the number of transactions within a certain period (such as a month or a quarter), identify users who frequently trade high-risk products.

[0028] Analyze the fluctuation of the transaction amount, calculate the coefficient of variation of the transaction amount (standard deviation divided by the mean). When the coefficient of variation exceeds a certain threshold, it is considered that there is an abnormal fluctuation in the transaction amount, and this is used as a characteristic of abnormal transaction behavior.

[0029] Consider the characteristics of the user's geographical location information Determine the user's geographical location based on the user's registered address or the IP address at the time of transaction.

[0030] Analyze the transaction risk characteristics of different regions. For example, some regions may be high-incidence areas of illegal wildlife trading, or some regions are close to wildlife habitats, and their transaction risks may be relatively high. By counting the number of illegal transaction cases in different regions, assign a risk weight to each region.

[0031] III. Feature Selection Statistical Analysis Calculate the correlation coefficient (such as Pearson correlation coefficient) between each feature and the risk of illegal transactions. For the commodity price feature, if a certain positive correlation is found between high - price commodities and the transactions of illegal wildlife products (because illegal products often have higher prices due to scarcity), then this feature may be considered a risk - related feature.

[0032] Information gain method Adopt the information gain algorithm to calculate the information gain value of each feature for classifying illegal transactions and legal transactions. For example, for the feature of "whether it contains keywords related to wildlife", calculate its information gain in the known samples of illegal transactions and legal transactions. If the information gain value is large, it indicates that this feature can effectively distinguish illegal transactions and legal transactions, and thus is selected as a feature related to the risk of illegal transactions.

[0033] IV. SVM Model Training Kernel Function Selection and Data Mapping Select a suitable kernel function, such as a linear kernel function, a polynomial kernel function, or a Gaussian kernel function. For this project, since the data may have non - linear relationships, initially try the Gaussian kernel function.

[0034] Use the selected kernel function to map the training data into a high - dimensional space. For the Gaussian kernel function (K(x,y)=\exp(-\gamma||x - y||^2)), where (\gamma) is the kernel function parameter, optimize the mapping effect of the data in the high - dimensional space by adjusting the value of (\gamma).

[0035] Cross - Validation and Parameter Tuning Adopt the k - fold cross - validation method (such as k = 5) to evaluate the model performance. Divide the training data set into 5 parts, each time select 4 parts as the training set and 1 part as the validation set, loop 5 times, and calculate evaluation metrics such as classification accuracy and recall rate each time.

[0036] During the parameter tuning process, adjust the parameters of the SVM model, the penalty parameter (C) (used to control the complexity of the model and prevent overfitting) and the kernel function parameter (\gamma). Through the method of grid search or random search, within a certain parameter range ((C\in[0.1,10]), (\gamma\in[0.01,1])), find the optimal parameter combination to make the performance of the model best in cross - validation.

[0037] Data Normalization Processing Normalize the training data using the min-max normalization method. For each feature (x_i), it is transformed into (x_{i}^{'}=\frac{x_i - min(x)}{max(x)-min(x)}), where (min(x)) and (max(x)) are the minimum and maximum values of the feature respectively. This ensures that each feature has the same weight in model training and avoids some features having too much influence on the model due to their large numerical ranges.

[0038] Introduce a regularization term Introduce the regularization term (\frac{1}{2}||w||^2) (where (w) is the weight vector of the model) into the objective function of the SVM model, and control the strength of regularization by adjusting the penalty parameter (C). When (C) is small, the role of the regularization term is large, and the model will tend to choose a simple decision boundary, thus preventing the model from overfitting.

[0039] In addition, adopt the ensemble learning method to combine multiple SVM models. Use the Bagging or Boosting method to build the ensemble model. For the Bagging method, multiple sub-datasets can be randomly sampled from the training set with replacement, and multiple SVM models are trained respectively. Then, the prediction results of these models are averaged or voted to obtain the final prediction result. For the Boosting method, the weights of the training samples can be adjusted according to the prediction errors of the previous model, and multiple SVM models are gradually trained. Finally, the results of these models are combined to improve the stability and accuracy of the model.

[0040] V. Risk Assessment Set the risk threshold Set a reasonable risk threshold according to historical data and actual regulatory requirements. Through the analysis of the model output probabilities of known illegal transaction cases, it is found that when the probability that the transaction output by the model belongs to the illegal transaction category exceeds 0.8, the transaction is highly likely to be an illegal transaction. Therefore, 0.8 is used as the threshold for high-risk transactions.

[0041] Risk assessment calculation Apply the trained SVM model to real-time transaction data. For each transaction, the model calculates the probability that it belongs to the illegal transaction category based on the extracted features. When a transaction contains multiple features related to illegal wildlife trading (such as the product description contains keywords of illegal wildlife products, the user has a history of frequently trading high-risk goods, etc.), the probability output by the model may be high.

[0042] Risk level judgment Judge the risk level of transactions according to the set threshold. If the probability output by the model is lower than the threshold (e.g., 0.2), mark the transaction as a low-risk transaction; if the probability is between the thresholds (e.g., 0.2 - 0.8), mark it as a medium-risk transaction; if the probability exceeds the threshold (e.g., 0.8), mark it as a high-risk transaction.

[0043] VI. Regulatory Decision Support Marking High-Risk Transactions and Automatic Alarm In the transaction monitoring system of the e-commerce platform, for transactions marked as high-risk, the system automatically marks and highlights them. Meanwhile, develop an alarm system that sends alarm information to the regulatory agency in real time when high-risk transactions are detected. The alarm information can include detailed transaction information (such as product information, user information, transaction time, etc.) so that the regulatory agency can take timely actions.

[0044] Providing Visualization Tools Develop visualization tools to display the risk assessment results and transaction trends using data visualization libraries (such as Tableau or D3.js). Through charts (such as bar charts showing the distribution of the number of transactions in different risk levels and line charts showing the trends of illegal transactions over a period of time), the regulatory agency can intuitively understand the situation of illegal wildlife transactions on the e-commerce platform and assist in formulating regulatory strategies.

[0045] Establishing a Risk Early Warning Mechanism Based on historical transaction data and model prediction results, establish a risk early warning mechanism. By analyzing the growth trend of high-risk transactions over a period of time, when it is found that the number of high-risk transactions continuously increases and the growth rate exceeds a certain proportion (e.g., 20%), issue a warning of an upward trend in potential illegal transactions. At the same time, customize and set risk assessment indicators and thresholds according to different regulatory requirements and scenarios. For the supervision of specific regions or specific product categories, risk assessment indicators (such as transaction amount thresholds for specific regions, keyword weights for specific products, etc.) and risk thresholds can be set separately to improve the pertinence and effectiveness of supervision.

[0046] VII. Model Update Updating Data Regularly Regularly (such as monthly or quarterly) obtain new illegal transaction cases and transaction data from the e-commerce platform.

[0047] Retraining the Model Use the new data to retrain the SVM model. During the retraining process, repeat the above steps of feature selection, model training, etc., and adjust the parameters and structure of the model to improve the accuracy and adaptability of the model. If new illegal wildlife transaction patterns (such as new types of illegal products or new transaction means) appear in the new data, through retraining the model, enable it to identify these new risk features.

[0048] VIII. Data Security Protection Encryption Measures The collected transaction data is encrypted using an encryption algorithm (such as the AES symmetric encryption algorithm). During data storage and transmission, the data is converted into ciphertext form, and only authorized personnel or systems with the correct key can decrypt and access the data.

[0049] Access Control A strict access control mechanism is established. Different access permissions are set for users at different levels (such as data administrators, analysts, supervisors, etc.). The data administrator has the highest authority and can perform operations such as adding, deleting, modifying, and querying data; analysts can only access desensitized data for model training and analysis; supervisors can only view risk assessment results and relevant transaction information summaries and cannot directly access the original transaction data. In this way, the security and confidentiality of the collected and processed transaction data are ensured.

[0050] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0051] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included within the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

[0052] In addition, it should be understood that although this specification is described according to embodiments, not every embodiment only contains an independent technical solution. This narrative way of the specification is only for clarity. Those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for mining regulatory risk data of illegal wildlife trading e-commerce platforms based on SVM, characterized in that: The following steps are involved: Data collection: collect transaction data from e-commerce platforms, including product information, user information, and transaction records; Data preprocessing: clean the collected data, remove duplicate, invalid or abnormal data, format it, and extract key features through natural language processing (NLP) and image processing technology; Feature selection: using statistical analysis and information gain methods to select features related to illegal transaction risks from preprocessed data; SVM model training: Use the selected features and known illegal transaction cases as training sets to train the SVM model. During the training process, kernel functions are used to map data into high-dimensional space to find the best classification hyperplane. Through cross-validation and parameter tuning, the generalization ability and classification accuracy of the model are improved. Risk assessment: Apply the trained SVM model to real-time transaction data to conduct risk assessment on each transaction. The model will calculate the probability that the transaction belongs to the illegal transaction category based on the extracted features, and determine the risk level of the transaction based on the set threshold; Regulatory decision support: Based on risk assessment results, provide decision support to regulators, including marking high-risk transactions, automatic alarms, and recommended regulatory measures; Model update: Regularly update the SVM model based on new illegal transaction cases and transaction data to improve the accuracy and adaptability of the model; Data security protection: Take measures such as encryption and access control to ensure the security and confidentiality of collected and processed transaction data.

2. According to claim 1, a method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM, characterized in that: The feature extraction comprises: Use NLP technology to extract keywords and phrases from product descriptions and identify words related to wildlife; Use image processing technology to identify species characteristics in product images, such as shape, color, texture, etc. Analyze users' historical transaction records and extract abnormal transaction behavior characteristics, such as frequent transactions of high-risk commodities and abnormal fluctuations in transaction amounts; Consider the user's geographic location information and analyze the transaction risk characteristics of different regions.

3. The method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM according to claim 1, characterized in that: The SVM model construction includes: Select appropriate kernel functions and parameters to map training data into high-dimensional space; The cross-validation method is used to evaluate the model performance and select the optimal kernel function and parameter combination; Normalize the training data to ensure that each feature has the same weight in model training; Regularization terms are introduced to prevent the model from overfitting.

4. According to claim 1, a method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM, characterized in that: The risk assessment and decision support include: Set a reasonable risk threshold and determine the risk level of the transaction based on the probability value output by the model; Develop an alarm system to monitor high-risk transactions in real time and automatically send alarm information; Provide visualization tools to display risk assessment results and transaction trends, and assist regulators in formulating regulatory strategies; Establish a risk warning mechanism to predict and warn of potential illegal transaction trends.

5. According to claim 2, a method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM, characterized in that: The feature extraction also uses deep learning technology to perform more accurate species identification and feature extraction on product images.

6. The method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM according to claim 3 is characterized by: The SVM model construction also adopts an integrated learning method to combine multiple SVM models to improve the stability and accuracy of the model.

7. The method for mining regulatory risk data of an e-commerce platform for illegal wildlife trade based on SVM according to claim 4, characterized in that: The risk assessment and decision support also customizes risk assessment indicators and thresholds according to different regulatory needs and scenarios.

Images