Enterprise credit data anomaly detection method and device based on artificial intelligence

Abstract

An embodiment of the present application provides an artificial intelligence-based enterprise credit data anomaly detection method and device, the method comprising: collecting enterprise credit data and preprocessing the data; adaptively selecting an anomaly detection algorithm based on the type of enterprise credit data; training and validating a machine learning model; using the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner; establishing a feedback mechanism to compare the detection results with set data to optimize the model, and regularly updating the data and model; the machine learning model can learn the actual distribution of the data through training data, rather than just assuming a normal distribution. Even if the data distribution is skewed or multimodal, the model can more accurately identify outliers. Adding abnormal samples to train the model during the training process can improve the model's performance when facing outliers in real-world data.

Description

Technical Field

[0001] The present application relates to the technical field of credit data anomaly detection, and more specifically, to an artificial intelligence-based method and device for detecting anomalies in enterprise credit data. Background Art

[0002] Anomaly detection in corporate credit data involves monitoring and analyzing a company's credit data using statistical analysis, machine learning, deep learning, and other techniques to identify anomalies that deviate from normal behavior patterns. These anomalies may indicate data entry errors, fraud, system vulnerabilities, or other risks. Anomaly detection is crucial for risk management, credit ratings, and compliance audits.

[0003] The 3-sigma rule is a common method for detecting anomalies in corporate credit data. This method is suitable for scenarios where the data is assumed to be roughly normally distributed. In corporate credit data analysis, this method can be used to quickly screen for possible input errors, fraud, or other unusual activity. For example, if a company's financial indicators suddenly exceed the 3-sigma range of long-term historical data, this may indicate potential risks or problems.

[0004] However, this method has some limitations: Dependence on normal distribution: The standard deviation method assumes that the data follows a normal distribution, which means that the data should be symmetrically distributed around the mean. However, data in the real world often do not conform to the normal distribution and may be skewed or have multiple peaks. In this case, the standard deviation method may not accurately reflect the true degree of discreteness of the data; Sensitivity to outliers: The standard deviation method is very sensitive to outliers. In a data set, even if there are only a few outliers, it may cause the standard deviation to increase significantly, thereby mistakenly marking normal data points as abnormal. Summary of the Invention

[0005] The purpose of the embodiments of the present application is to provide an artificial intelligence-based enterprise credit data anomaly detection method and device to solve the problem that existing methods cannot accurately reflect the true discreteness of the data and are very sensitive to outliers, resulting in normal data points being mistakenly marked as anomalies.

[0006] In a first aspect, the embodiments of the present application provide an artificial intelligence-based method for detecting anomalies in enterprise credit data, including:

[0007] Collect enterprise credit data and pre-process the data;

[0008] Adaptively select anomaly detection algorithms based on the type of enterprise credit data;

[0009] Train and validate machine learning models;

[0010] Use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner;

[0011] Establish a feedback mechanism to compare the test results with the set data to optimize the model, and update the data and model regularly.

[0012] In the above implementation process, corporate credit data is collected and preprocessed; an anomaly detection algorithm is adaptively selected according to the type of corporate credit data; a machine learning model is trained and verified; the trained model is used to perform real-time detection on newly input credit data, anomalies are identified, and the detection results are presented in a visual manner; a feedback mechanism is established to compare the detection results with the set data to optimize the model, and the data and model are updated regularly; the machine learning model can learn the actual distribution of the data through training data, rather than just assuming a normal distribution. Even if the data distribution is skewed or multimodal, the model can more accurately identify outliers. Adding abnormal samples to train the model during the training process can improve the model's performance when facing outliers in real-world data.

[0013] Furthermore, the collection of enterprise credit data and pre-processing of the data include:

[0014] Collect corporate credit data and perform data cleaning on it; this data cleaning includes processing missing values, removing duplicate data and outliers;

[0015] Select key features of corporate credit data and perform normalization and standardization.

[0016] In the above implementation process, the enterprise credit data is preprocessed to ensure the quality of the data.

[0017] Furthermore, the adaptive selection of anomaly detection algorithms based on the enterprise credit data type includes:

[0018] Analyze corporate credit data types;

[0019] According to the enterprise credit data type, the anomaly detection algorithm is adaptively selected according to the algorithm selection strategy.

[0020] In the above implementation process, the appropriate anomaly detection algorithm is automatically selected according to the type of enterprise credit data to improve the accuracy and efficiency of anomaly detection.

[0021] Furthermore, the training and verification of the machine learning model includes:

[0022] Obtain training data and divide the data set into training set and test set;

[0023] Adjust the hyperparameters of the machine learning model, including cross-validation and grid search.

[0024] In the above implementation process, the dataset is divided into training set and test set to ensure the generalization ability of the model; the hyperparameters of the model are adjusted to optimize the performance of the model.

[0025] Furthermore, the collection of enterprise credit data and data cleaning of the data include:

[0026] Conduct a comprehensive check on the enterprise credit data to identify the fields with missing values ​​and record the number of missing values ​​and the fields in which they are located;

[0027] Analyze the reasons for missing values ​​and classify missing values ​​into missing types according to the reasons for missing values. Missing types include: random missing, completely random missing, and non-random missing.

[0028] Select missing value processing method based on the proportion of missing values;

[0029] Use set functions to identify duplicate data in corporate credit data and record the number and specific content of duplicate data;

[0030] Delete or merge duplicate data;

[0031] Identify outliers in corporate credit data and analyze the causes of outliers;

[0032] Anomaly handling methods are used to handle outliers, including deleting outliers, replacing outliers, and binning.

[0033] The selection of key features of enterprise credit data and normalization and standardization processing include:

[0034] Select key features based on feature selection methods, including filtering methods, wrapper methods, and embedded methods;

[0035] For each selected key feature, determine its normalization method;

[0036] Calculate statistics of key features;

[0037] Transform the eigenvalues ​​according to the selected normalization method;

[0038] Select a standardization method;

[0039] Calculate the mean and standard deviation of each key feature;

[0040] For each eigenvalue, subtract the mean of the key feature from it, and then divide it by the standard deviation of the key feature to obtain the standardized eigenvalue.

[0041] Furthermore, the method of adaptively selecting an anomaly detection algorithm based on the enterprise credit data type and the algorithm selection strategy includes:

[0042] For numerical data, choose statistical methods or machine learning methods;

[0043] For data with time series characteristics, choose time series analysis methods or deep learning methods;

[0044] For public opinion data, choose text processing and machine learning methods;

[0045] The algorithm selection strategy includes:

[0046] Perform feature analysis on the input corporate credit data to determine the data type, distribution characteristics, and dimensions, and calculate the statistical characteristics of the data;

[0047] Build a library of anomaly detection algorithms targeting different data types;

[0048] Encapsulate each algorithm to give it a unified interface;

[0049] Select evaluation metrics to measure the performance of the algorithm, including accuracy, recall, and F1 value;

[0050] For different types of data, choose different evaluation indicators;

[0051] According to the results of feature analysis, select an algorithm suitable for the current data from the algorithm library;

[0052] Use a portion of labeled data sets to evaluate the algorithm and select the algorithm based on the evaluation indicators;

[0053] If the data type changes or new data is added, re-analysis of features and selection of algorithms are performed to make adaptive adjustments.

[0054] Furthermore, adjusting the hyperparameters of the machine learning model includes:

[0055] Select a cross-validation method, where the cross-validation methods include: K-fold cross-validation, leave-one-out cross-validation, and stratified cross-validation;

[0056] Divide the enterprise credit data set according to the selected cross-validation method;

[0057] For each split, the model is trained using the training set and the performance indicators of the model are evaluated on the validation set, including accuracy, recall and F1 value;

[0058] Record the results of each validation and calculate the mean and standard deviation to evaluate model performance;

[0059] For anomaly detection models, determine the hyperparameters that need to be tuned;

[0060] Combine the value range of each hyperparameter into a parameter grid;

[0061] Use cross-validation to train and validate the model on each hyperparameter combination in the parameter grid;

[0062] Record the verification results of each hyperparameter combination and select the hyperparameter combination;

[0063] If the results obtained by grid search do not meet the set requirements, search for hyperparameter combinations.

[0064] In a second aspect, an embodiment of the present application provides an artificial intelligence-based enterprise credit data anomaly detection device, comprising:

[0065] Data collection module, used to collect enterprise credit data and pre-process the data;

[0066] Algorithm selection module, used to adaptively select anomaly detection algorithm based on the enterprise credit data type;

[0067] Model training module, used to train and validate machine learning models;

[0068] The anomaly detection module is used to use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner;

[0069] The feedback adjustment module is used to establish a feedback mechanism, compare the test results with the set data to optimize the model, and regularly update the data and model.

[0070] In a third aspect, an embodiment of the present application provides an electronic device, including:

[0071] A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the above-mentioned artificial intelligence-based enterprise credit data anomaly detection method.

[0072] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a server, the method for detecting anomalies in enterprise credit data based on artificial intelligence as described above is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0074] Figure 1 A flowchart of an artificial intelligence-based enterprise credit data anomaly detection method provided in an embodiment of the present application;

[0075] Figure 2 This is a schematic diagram of the structure of an enterprise credit data anomaly detection device based on artificial intelligence provided by an embodiment of the present application;

[0076] Figure 3 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0078] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0079] Anomaly detection in corporate credit data involves monitoring and analyzing a company's credit data using statistical analysis, machine learning, deep learning, and other techniques to identify anomalies that deviate from normal behavior patterns. These anomalies may indicate data entry errors, fraud, system vulnerabilities, or other risks. Anomaly detection is crucial for risk management, credit ratings, and compliance audits.

[0080] The 3-sigma rule is a commonly used method for detecting anomalies in corporate credit data. This method, also known as the 3-sigma rule or the Laida criterion, is a statistical method based on the normal distribution assumption and is used to identify outliers in a dataset. The core idea of ​​this method is that in a normal distribution, the vast majority of data points are clustered within a region around the mean, while data points far from the mean are likely outliers.

[0081] Calculate the mean and standard deviation: First, calculate the mean (μ) and standard deviation (σ) of the given corporate credit data set. These statistics describe the central location of the data and the degree of dispersion of the data distribution, respectively.

[0082] Applying the 3-sigma rule: According to the characteristics of the normal distribution, approximately 99.73% of the data points will fall within three standard deviations of the mean, that is, (μ-3σ,μ+3σ). Therefore, the standard deviation method considers data points that are more than three standard deviations away from the mean as outliers.

[0083] Identify outliers: Calculate the difference between each data point and the mean and check whether these differences exceed the ±3σ limit. Data points outside this range will be marked as outliers.

[0084] The standard deviation method is suitable for scenarios where the data is assumed to be roughly normally distributed. In corporate credit data analysis, this method can be used to quickly screen for possible input errors, fraud, or other unusual activity. For example, if a company's financial indicators suddenly fall outside the 3-sigma range of long-term historical data, this may indicate potential risks or problems.

[0085] The standard deviation method is commonly used in anomaly detection of corporate credit data, but this method has some limitations: Dependence on normal distribution: The standard deviation method assumes that the data follows a normal distribution, which means that the data should be symmetrically distributed around the mean. However, data in the real world often does not conform to a normal distribution and may be skewed or have multiple peaks. In this case, the standard deviation method may not accurately reflect the true degree of discreteness of the data; Sensitivity to outliers: The standard deviation method is very sensitive to outliers. In a data set, even if there are only a few outliers, it may cause the standard deviation to increase significantly, thereby mistakenly marking normal data points as anomalies.

[0086] Based on this, an embodiment of the present application proposes an enterprise credit data anomaly detection method based on artificial intelligence to solve the above problems.

[0087] Please see Figure 1 , Figure 1 This is a flow chart of an AI-based method for detecting anomalies in enterprise credit data provided by an embodiment of the present application. The AI-based method for detecting anomalies in enterprise credit data includes:

[0088] 100. Collect enterprise credit data and pre-process the data.

[0089] 110. Collect corporate credit data and perform data cleaning on the data; data cleaning includes processing missing values, removing duplicate data and outliers.

[0090] Optionally, corporate credit data includes financial statements, transaction records, customer feedback, and social media data.

[0091] 111. Conduct a comprehensive check on the enterprise credit data, identify the fields with missing values, and record the number of missing values ​​and the fields in which they are located.

[0092] Specifically, first, conduct a comprehensive check on the corporate credit data to determine which fields have missing values. You can use data visualization tools (such as bar charts, heat maps, etc.) or data analysis software to quickly identify the distribution of missing values; record the number of missing values ​​and the fields in which they are located so that targeted processing measures can be taken later.

[0093] 112. Analyze the reasons for missing values ​​and classify missing values ​​into missing types based on the reasons for missing values. Missing types include: random missing, completely random missing, and non-random missing.

[0094] Specifically, understanding the causes of missing values ​​is crucial for selecting appropriate treatment methods. Missing values ​​may be caused by errors in the data collection process, data transmission problems, or the surveyed enterprise's failure to provide information. Alternatively, depending on the cause of missing values, missing values ​​can be categorized as missing at random, missing completely at random, or missing not at random.

[0095] 113. Select the missing value handling method based on the proportion of missing values.

[0096] Alternatively, if the proportion of missing values ​​is low, you can consider directly deleting records containing missing values. This method is simple and direct, but it may result in a reduction in the amount of data and affect the accuracy of subsequent analysis. In addition, for a specific variable, if the proportion of missing values ​​is high and the variable is less important for credit assessment, you can consider deleting the variable.

[0097] Optionally, for numeric variables, you can use the mean or median of the variable to fill missing values. This method is suitable for symmetrical data distributions. For categorical variables, you can use the mode of the variable to fill missing values. Additionally, you can use linear interpolation, polynomial interpolation, and other methods to fill missing values. This method is suitable for data with certain trends or patterns.

[0098] Optionally, you can also use model prediction to fill in missing values, using machine learning algorithms such as regression models and decision trees. This method can make full use of other information in the data, but the computational cost is high.

[0099] 114. Use set functions to identify duplicate data in corporate credit data and record the number and specific content of the duplicate data.

[0100] Specifically, data processing tools or functions in programming languages ​​are used to deduplicate corporate credit data. The presence of duplicate data can be determined by comparing key fields of each record in the data set (such as company name, unified social credit code, etc.), and the number and specific content of the duplicate data can be recorded for subsequent processing.

[0101] 115. Delete or merge duplicate data.

[0102] Specifically, delete duplicate records: For completely duplicate records, you can directly delete the redundant records and only keep one unique record.

[0103] Specifically, merging duplicate records: If there are some differences between duplicate records, you can merge them based on the specific situation. For example, you can merge different field values ​​in duplicate records, or select one record as a representative and integrate the information of other records into it.

[0104] 116. Identify outliers in corporate credit data and analyze the causes of the outliers.

[0105] Specifically, outliers refer to data points that are significantly different from other data in the data set. The following methods can be used to identify outliers: Statistical methods: By calculating statistical quantities such as the mean, median, and standard deviation of the data, and using methods such as box plots and the 3σ principle to determine outliers, box plots can intuitively display the distribution of data. Outliers are usually defined as data points outside the upper and lower boundaries of the box plot. The 3σ principle is based on the assumption of normal distribution and considers data points that are more than 3 times the standard deviation away from the mean as outliers; Machine learning methods: Using unsupervised learning algorithms, such as clustering algorithms and isolation forests, to identify outliers. These algorithms can automatically discover abnormal patterns in data without pre-setting the definition of outliers.

[0106] Understandably, determining the cause of outliers is very important, as some outliers may be due to data errors or unusual circumstances, while others may reflect real special circumstances.

[0107] Optionally, for outliers that may be caused by data errors, further inspection of the data collection and processing process is required to determine whether there are errors; for outliers that reflect real special circumstances, consideration may be given to retaining these data points, but they require special processing in subsequent analysis.

[0108] 117. Use an exception handling method to handle outliers, wherein the exception handling method includes deleting outliers, replacing outliers, and binning.

[0109] For example, deleting outliers: If the number of outliers is small, you can consider deleting these outliers directly.

[0110] For example, replacing outliers: You can use statistics such as the mean, median, and mode to replace outliers. This method is suitable for situations where the number of outliers is small and their impact on the data distribution is small.

[0111] For example, binning: For continuous variables, you can bin them and assign outliers to specific bins. This method can reduce the impact of outliers on the model while preserving the distribution information of the data.

[0112] As mentioned above, when dealing with missing values, removing duplicate data, and removing outliers, continuous data quality checks and verification are necessary to ensure that the processed data is accurate, complete, and reliable. At the same time, appropriate processing methods can be selected based on specific business needs and data analysis objectives to improve data quality and usability.

[0113] 120. Select key features of corporate credit data and perform normalization and standardization.

[0114] Understandably, clarifying the specific objectives of a company's credit assessment, such as assessing its solvency, operational stability, and regulatory compliance, helps determine which data characteristics are most relevant to the credit assessment. Additionally, communicate with business departments to understand the key indicators and risk points they focus on when assessing corporate credit.

[0115] Additionally, research the relevant theories and methods of corporate credit assessment and gain an understanding of traditional credit assessment indicators, such as financial ratios (debt-to-asset ratio, current ratio, etc.), operating performance indicators (revenue growth rate, net profit margin, etc.), and industry-specific indicators. Consider the characteristics of different industries and identify key characteristics applicable to specific sectors. For example, for manufacturing companies, characteristics such as production capacity and inventory turnover may be of particular concern; for service companies, customer satisfaction and employee turnover rates may be more important.

[0116] Optionally, collect data sets related to corporate creditworthiness, including financial data, operational data, market data, and public opinion data. Perform data visualization and statistical analysis to understand the distribution and correlation of various features, as well as their potential relationship to credit assessment objectives. For example, observe the distribution of features by plotting histograms and scatter plots, and calculate correlation coefficients between features to identify redundant features.

[0117] 121. Select key features based on feature selection methods, where feature selection methods include filtering methods, wrapping methods, and embedded methods.

[0118] Exemplary filtering method: selection is made based on the correlation between features and target variables. Indicators such as the Pearson correlation coefficient and mutual information can be used to measure the degree of correlation between features and credit assessment results, and features with higher correlation are selected.

[0119] An exemplary, wrapper approach is to treat feature selection as an optimization problem and use machine learning algorithms (such as random forests, support vector machines, etc.) to perform feature selection. By continuously trying different feature combinations, evaluating model performance, and selecting the optimal feature subset.

[0120] An example of an embedded approach is to automatically perform feature selection during model training. For example, when using Lasso or Ridge regression, the regularization term can encourage the model to automatically select important features and compress the coefficients of unimportant features.

[0121] 122. For each selected key feature, determine its normalization method.

[0122] For example, minimum-maximum normalization: map the eigenvalues ​​to the interval [0, 1], and the calculation formula is:

[0123]

[0124] Among them, X is the original eigenvalue, X min is the minimum value of the feature, X max is the maximum value of the feature.

[0125] For example, mean normalization: maps the eigenvalues ​​to a specific interval centered on the mean, and the calculation formula is:

[0126]

[0127] Here, μ is the mean of the feature and σ is the standard deviation of the feature.

[0128] For each selected feature, determine its normalization method.

[0129] 123. Calculate statistics of key features.

[0130] Specifically, statistics such as the minimum, maximum, mean, and standard deviation of the features are calculated.

[0131] 124. Transform the eigenvalues ​​according to the selected normalization method.

[0132] Optionally, the eigenvalues ​​are transformed according to the selected normalization method. For example, if min-max normalization is used, for each eigenvalue, the original eigenvalue is subtracted from the minimum value of the feature, and then divided by the difference between the maximum and minimum values ​​of the feature to obtain the normalized eigenvalue.

[0133] 125. Select a standardization method.

[0134] For example, Z-score standardization converts the eigenvalues ​​into a distribution with a mean of 0 and a standard deviation of 1. The calculation formula is:

[0135]

[0136] Here, μ is the mean of the feature and σ is the standard deviation of the feature.

[0137] 126. Calculate the mean and standard deviation of each key feature.

[0138] 127. For each eigenvalue, subtract the mean of the key feature from it, and then divide it by the standard deviation of the key feature to obtain the standardized eigenvalue.

[0139] As described above, the embodiments of the present application can select key features and perform normalization and standardization processing on them, providing high-quality input data for artificial intelligence-based enterprise credit data anomaly detection and improving the performance and accuracy of the model.

[0140] 200. Adaptively select anomaly detection algorithm based on the type of enterprise credit data.

[0141] 210. Analyze corporate credit data types.

[0142] For example, financial data typically includes data such as balance sheets, income statements, and cash flow statements. These data are numerical and may exhibit certain distribution patterns. For example, indicators such as debt-to-asset ratios, current ratios, and net profits may be included.

[0143] For example, operational data covers a company's production, sales, inventory, and other aspects. It may contain both numerical and categorical features, and the data distribution may be complex. For example, sales, output, customer satisfaction ratings, etc.

[0144] For example, market data has the following characteristics: it reflects the performance of enterprises in the market, such as stock price, market capitalization, price-to-earnings ratio, etc. The data may have time series characteristics and be significantly affected by market factors, such as stock price fluctuations and market share changes.

[0145] For example, public opinion data is mainly text data, including news reports, social media comments, etc., which requires text processing and analysis to extract valuable information. For example, news headlines, Weibo comments, etc.

[0146] 220. Based on the enterprise credit data type, the anomaly detection algorithm is adaptively selected according to the algorithm selection strategy.

[0147] 221. For numerical data, choose statistical methods or machine learning methods.

[0148] Exemplary statistical methods: Normal distribution hypothesis test: If the data approximately follows a normal distribution, outliers can be detected using methods such as the 3σ principle or the Grubbs test. These methods calculate a threshold based on the mean and standard deviation of the data, and data points exceeding the threshold are considered outliers.

[0149] An exemplary statistical method is as follows: Box plot method: By drawing a box plot, the distribution of the data is observed, and data points that are beyond 1.5 times the interquartile range of the upper and lower quartiles are regarded as outliers.

[0150] Exemplary machine learning methods: Isolation Forest: Applicable to high-dimensional data, it can effectively detect outliers. This algorithm quickly isolates outliers by randomly splitting the data space.

[0151] Exemplary machine learning method: Local Outlier Factor (LOF): Calculate the local outlier factor of each data point. The larger the value, the more likely the point is an outlier. It takes into account the local density of the data point and works better for data with different density areas.

[0152] 222. For data with time series characteristics, choose time series analysis methods or deep learning methods.

[0153] For example, the time series analysis method is: ARIMA model residual analysis: an autoregressive moving average model (ARIMA) is established to fit the time series data, and then the model residuals are analyzed to see if there are outliers. If the residuals exceed a certain range, the corresponding time point data can be considered as an outlier.

[0154] Exemplary time series analysis method: Sliding window method: Set a sliding window of fixed length and calculate the statistical characteristics of the data in the window, such as mean, standard deviation, etc. When the statistical characteristics of the new data point are significantly different from those of the data in the window, it is judged as an outlier.

[0155] Exemplary deep learning methods: Long Short-Term Memory Network (LSTM): can model time series data and detect anomalies by predicting the difference between future values ​​and actual values. If the prediction error is large, there may be anomalies.

[0156] 223. For public opinion data, choose text processing and machine learning methods.

[0157] For example, text preprocessing: cleaning, word segmentation, and stop word removal of public opinion data to convert the text into a computable vector representation.

[0158] For example, topic models, such as Latent Dirichlet Allocation (LDA), can discover the topic distribution in a text. Abnormal public opinion may differ greatly from common topics, and whether it is abnormal can be determined by calculating the distance between the text and the topic.

[0159] For example, a machine learning classifier uses classifiers such as support vector machines (SVM) and random forests to classify public opinion data into two categories: normal and abnormal. It is trained by labeling a portion of the data, and then new public opinion data is classified.

[0160] Based on the above embodiment, the method of the embodiment of the present application can also be concretized as follows: the algorithm selection strategy includes:

[0161] 231. Perform feature analysis on the input corporate credit data to determine the data type, distribution characteristics, and dimensions, and calculate the statistical characteristics of the data.

[0162] Specifically, the feature analysis module is used to perform feature analysis on the input corporate credit data to determine the data type, distribution characteristics, dimensions and other information; the statistical characteristics of the data, such as mean, standard deviation, skewness, kurtosis, etc., are calculated to provide a basis for algorithm selection.

[0163] 232. Build a library containing multiple anomaly detection algorithms, targeting different data types.

[0164] Optionally, build a library of multiple anomaly detection algorithms, including the above algorithms for different data types.

[0165] 233. Encapsulate each algorithm to give it a unified interface.

[0166] It is understandable that each algorithm is encapsulated to have a unified interface for easy calling and evaluation.

[0167] 234. Select evaluation indicators to measure the performance of the algorithm, including: accuracy, recall rate, and F1 value.

[0168] 235. Choose different evaluation indicators for different types of data.

[0169] Optionally, different evaluation metrics can be selected for different types of data based on their characteristics. For example, for financial data, accuracy is more important, while for public opinion data, recall may be more important.

[0170] 236. Based on the results of feature analysis, select an algorithm suitable for the current data from the algorithm library.

[0171] 237. Use a portion of labeled data sets to evaluate the algorithm and select the algorithm based on the evaluation indicators.

[0172] 238. If the data type changes or new data is added, re-analyze the features and select the algorithm to make adaptive adjustments.

[0173] As described above, the embodiment of the present application automatically selects a suitable anomaly detection algorithm according to the type of enterprise credit data, thereby improving the accuracy and efficiency of anomaly detection.

[0174] 300. Train and validate machine learning models.

[0175] 310. Obtain training data and divide the dataset into training set and test set.

[0176] 320. Adjust the hyperparameters of the machine learning model, where the hyperparameters include: cross validation and grid search.

[0177] Optionally, understand the purpose of cross-validation. Cross-validation is mainly used to evaluate the performance and stability of the model. It can also help choose appropriate hyperparameters. It divides the dataset into multiple subsets and trains and validates on different subset combinations to obtain more reliable model evaluation results.

[0178] 321. Select a cross-validation method, where the cross-validation method includes: K-fold cross-validation, leave-one-out cross-validation, and stratified cross-validation.

[0179] For example, K-fold cross-validation: randomly divide the dataset into K subsets of similar size, select one of the subsets as the validation set, and the remaining K-1 subsets as the training set, perform K training and validation, and finally calculate the average of the K validation results as the performance indicator of the model.

[0180] For example, leave-one-out cross-validation: When the dataset is small, you can use leave-one-out cross-validation. Each time, only one sample is left as the validation set, and the remaining samples are used as the training set. Training and validation are performed N times (N is the size of the dataset).

[0181] For example, stratified cross-validation: In classification problems, if the category distribution of the dataset is unbalanced, stratified cross-validation can be used to ensure that the category ratio in each subset is the same as that in the original dataset.

[0182] 322. Divide the enterprise credit data set according to the selected cross-validation method.

[0183] 323. For each split, the model is trained using the training set and the performance indicators of the model are evaluated on the validation set, including accuracy, recall and F1 value.

[0184] 324. Record the results of each validation and calculate the mean and standard deviation to evaluate the model performance.

[0185] 325. For anomaly detection models, determine the hyperparameters that need to be adjusted.

[0186] Optionally, determine the hyperparameters that need to be adjusted for the anomaly detection model. For example, for the isolation forest model, possible hyperparameters include the number of trees and subsampling size; for the support vector machine model, hyperparameters may include the kernel function type and regularization parameter. Each hyperparameter needs to be set to determine a reasonable range of values.

[0187] 326. Combine the value range of each hyperparameter into a parameter grid.

[0188] For example, the value range of each hyperparameter is combined into a parameter grid. For example, if there are two hyperparameters A and B, the value range of A is [1, 2, 3], and the value range of B is [0.1, 0.2, 0.3], then the parameter grid is {(1, 0.1), (1, 0.2), (1, 0.3), (2, 0.1), (2, 0.2), (2, 0.3), (3, 0.1), (3, 0.2), (3, 0.3)}.

[0189] 327. Use cross-validation to train and validate the model on each hyperparameter combination in the parameter grid.

[0190] 328. Record the verification results of each hyperparameter combination and select the hyperparameter combination.

[0191] Optionally, the best performing hyperparameter combination is selected. The best performance can be determined based on a specific evaluation metric, such as the highest accuracy, the largest F1 value, etc.

[0192] 329. If the results of the grid search do not meet the set requirements, search for the hyperparameter combination.

[0193] Optionally, if the grid search results are not satisfactory, you can perform a more refined search near the best-performing hyperparameter combination. For example, you can narrow the range of hyperparameter values ​​or use a denser grid for the search.

[0194] Optionally, you can also try other hyperparameter tuning methods, such as random search. Random search randomly selects some points in the hyperparameter space for evaluation. Compared with grid search, it can try more hyperparameter combinations under the same computing resources.

[0195] In addition, when adjusting hyperparameters, we should not only pay attention to the performance of the model on the validation set, but also consider the complexity and generalization ability of the model. An overly complex model may overfit the training data and perform poorly on new data.

[0196] For example, you can determine whether the model is overfitting by observing the performance difference between the training set and the validation set. If the performance on the training set is very good, but the performance on the validation set is poor, it means that the model may be overfitting. You can try to reduce the complexity of the model, such as reducing the number of trees, lowering the regularization parameter, etc.

[0197] Therefore, the final hyperparameter combination you choose should be based on actual business needs. For example, if you require a high recall rate for the model, you can choose a hyperparameter combination that performs well in terms of recall, even if this may reduce accuracy. Also, consider factors such as the model's computational efficiency and interpretability.

[0198] In some embodiments, model performance is evaluated using metrics such as accuracy, recall, and F1 score.

[0199] For example, accuracy refers to the ratio of the number of samples correctly predicted by the model to the total number of samples. The calculation formula is:

[0200]

[0201] Among them, TP (True Positive) represents true positive examples, that is, the number of samples that are actually positive examples and predicted as positive examples by the model; TN (True Negative) represents true negative examples, that is, the number of samples that are actually negative examples and predicted as negative examples by the model; FP (False Positive) represents false positive examples, that is, the number of samples that are actually negative examples but predicted as positive examples by the model; FN (False Negative) represents false negative examples, that is, the number of samples that are actually positive examples but predicted as negative examples by the model.

[0202] Specifically, for the corporate credit data anomaly detection model, abnormal companies in the data set are marked as positive examples, and normal companies are marked as negative examples; the model makes predictions for each company to determine whether it is an abnormal company; the number of true positive examples, true negative examples, false positive examples, and false negative examples is counted, and then the accuracy is calculated according to the formula; the higher the accuracy, the better the overall prediction accuracy of the model; however, when the ratio of positive and negative examples in the data set is unbalanced, the accuracy may be misleading.

[0203] For example, recall rate (Recall), also known as recall rate, refers to the ratio of the number of positive examples correctly predicted by the model to the total number of actual positive examples. The calculation formula is:

[0204]

[0205] Specifically, in the anomaly detection of corporate credit data, the recall rate measures how many abnormal companies the model can correctly identify. A higher recall rate means that the model can find all abnormal companies as much as possible and reduce the number of missed detections. However, simply pursuing a high recall rate may lead to an increase in false positives, that is, some normal companies are mistakenly identified as abnormal companies.

[0206] For example, the F1 score is the harmonic mean of precision and recall, which takes into account the accuracy and completeness of the model. The calculation formula is:

[0207]

[0208] in, is the precision, which indicates the proportion of true positive examples among the samples predicted by the model as positive examples.

[0209] Specifically, the F1 score strikes a balance between precision and recall. A higher F1 score indicates that the model performs well in both accuracy and completeness. When evaluating anomaly detection models for corporate credit data, the F1 score can be optimized by adjusting the model's hyperparameters or selecting different algorithms.

[0210] For example, a comprehensive evaluation is performed based on various indicators: ROC curve is drawn and AUC value is calculated.

[0211] The ROC (Receiver Operating Characteristic) curve is a curve drawn with the false positive rate (FPR) as the horizontal axis and the true positive rate (TPR, i.e., recall rate) as the vertical axis.

[0212] Among them, the AUC (Area Under the Curve) value is the area under the ROC curve. The larger the AUC value, the better the performance of the model.

[0213] Specifically, by drawing the ROC curve and calculating the AUC value, the performance of the model at different thresholds can be more comprehensively evaluated.

[0214] For example, the specific business needs of detecting anomalies in corporate credit data determine the importance of accuracy, recall, and F1 score. For example, if the risk of missing anomaly companies is high, recall may be more important; if the cost of misclassifying normal companies as anomalies is high, accuracy may be more important. Considering multiple indicators comprehensively, select the model that best suits the business needs.

[0215] As described above, the embodiments of the present application can comprehensively evaluate the performance of the enterprise credit data anomaly detection model by using indicators such as accuracy, recall rate, and F1 score, providing a basis for model optimization and improvement.

[0216] 400. Use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner.

[0217] Specifically, the trained model is used to perform real-time detection on newly input credit data to identify potential anomalies; a real-time monitoring system is built to present the detection results in a visual manner to facilitate timely processing by relevant personnel.

[0218] 500. Establish a feedback mechanism to compare test results with set data to optimize the model, and update data and models regularly.

[0219] For example, a feedback mechanism is established to compare the detection results with manually confirmed anomalies to optimize the model. Incremental learning: The model is updated regularly and retrained with new data to maintain its effectiveness.

[0220] In the above implementation process, the embodiment of the present application collects enterprise credit data and pre-processes the data; adaptively selects anomaly detection algorithms according to the type of enterprise credit data; trains and verifies machine learning models; uses the trained model to perform real-time detection on newly input credit data, identifies anomalies, and presents the detection results in a visual manner; establishes a feedback mechanism to compare the detection results with the set data to optimize the model, and regularly updates the data and model; the machine learning model can learn the actual distribution of data through training data, rather than just assuming a normal distribution. Even if the data distribution is skewed or multimodal, the model can more accurately identify outliers. Adding abnormal samples to train the model during the training process can improve the performance of the model when facing outliers in real-world data.

[0221] The machine learning model of the embodiment of the present application can learn the actual distribution of the data through training data, rather than just assuming a normal distribution, which means that even if the data distribution is skewed or multimodal, the model can more accurately identify outliers.

[0222] Deep learning models can be designed to be more robust to outliers. For example, adversarial training can make models more robust and less susceptible to outliers. Intentionally adding outliers to the training process can improve the model's performance in the presence of outliers in real-world data.

[0223] The traditional 3-sigma rule uses a fixed threshold to identify outliers. The machine learning model in the embodiment of the present application can dynamically adjust the threshold based on real-time changes in the data, thereby providing a more flexible and adaptable anomaly detection mechanism. This approach can better cope with situations where data distribution evolves over time.

[0224] The above steps are not to be performed in a strict order as described in the numbers, but should be understood as an overall solution.

[0225] In the second aspect, based on the above embodiments, Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based enterprise credit data anomaly detection device provided in an embodiment of the present application. Figure 2 The enterprise credit data anomaly detection device based on artificial intelligence provided in this embodiment specifically includes: a data collection module 201, an algorithm selection module 202, a model training module 203, an anomaly detection module 204 and a feedback adjustment module 205.

[0226] Among them, the data collection module 201 is used to collect enterprise credit data and pre-process the data; the algorithm selection module 202 is used to adaptively select anomaly detection algorithms based on the type of enterprise credit data; the model training module 203 is used to train and verify machine learning models; the anomaly detection module 204 is used to use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner; the feedback adjustment module 205 is used to establish a feedback mechanism, compare the detection results with the set data to optimize the model, and regularly update the data and model.

[0227] As described above, the embodiments of the present application collect enterprise credit data and pre-process the data; adaptively select anomaly detection algorithms based on the type of enterprise credit data; train and verify machine learning models; use the trained models to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner; establish a feedback mechanism to compare the detection results with the set data to optimize the model, and regularly update the data and model; the machine learning model can learn the actual distribution of data through training data, rather than just assuming a normal distribution. Even if the data distribution is skewed or multimodal, the model can more accurately identify outliers. Adding abnormal samples to train the model during the training process can improve the performance of the model when facing outliers in real-world data.

[0228] The artificial intelligence-based enterprise credit data anomaly detection device provided in the embodiment of the present application can be used to execute the artificial intelligence-based enterprise credit data anomaly detection method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0229] On the third aspect, an embodiment of the present application also provides an electronic device that can integrate the artificial intelligence-based enterprise credit data anomaly detection device provided in an embodiment of the present application. Figure 3 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device includes: an input device 33, an output device 34, a memory 32, and one or more processors 31; the memory 32 is used to store one or more programs; when the one or more programs are executed by the one or more processors 31, the one or more processors 31 implement the enterprise credit data anomaly detection method based on artificial intelligence as provided in the above embodiment. The input device 33, the output device 34, the memory 32, and the processor 31 can be connected by a bus or other means. Figure 3 The bus connection is taken as an example.

[0230] The processor 31 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 32, thereby realizing the above-mentioned artificial intelligence-based enterprise credit data anomaly detection method.

[0231] The electronic device provided above can be used to execute the artificial intelligence-based enterprise credit data anomaly detection method provided in the above embodiment, and has corresponding functions and beneficial effects.

[0232] In a fourth aspect, an embodiment of the present application also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned artificial intelligence-based enterprise credit data anomaly detection method, and can achieve the same beneficial effects.

[0233] Of course, the storage medium containing computer-executable instructions provided in an embodiment of the present application is not limited to the above-mentioned artificial intelligence-based enterprise credit data anomaly detection method, and can also execute related operations in the artificial intelligence-based enterprise credit data anomaly detection method provided in any embodiment of the present application.

[0234] In a fifth aspect, the embodiments of the present application further provide a computer program product, and the methods described in the various embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on a computer, the processes or functions described in the various embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model) or other programmable device.

[0235] The computer program or instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer program or instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless method. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; an optical medium, such as a digital video disk; or a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0236] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and the module, program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or can be implemented using a combination of dedicated hardware and computer instructions.

[0237] In addition, the functional modules in each embodiment of the present application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0238] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling an electronic device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0239] The foregoing is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included within the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following figures. Therefore, once an item is defined in one figure, it does not need to be further defined or explained in subsequent figures.

[0240] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0241] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

Claims

1. An artificial intelligence-based enterprise credit data anomaly detection method, characterized in that: The method comprises: Collect enterprise credit data and pre-process the data; Adaptively select anomaly detection algorithms based on the type of enterprise credit data; Train and validate machine learning models; Adjusting hyperparameters of machine learning models, including cross-validation and grid search; Select a cross-validation method, where the cross-validation methods include: K-fold cross-validation, leave-one-out cross-validation, and stratified cross-validation; Divide the enterprise credit data set according to the selected cross-validation method; For each split, the model is trained using the training set and the performance indicators of the model are evaluated on the validation set, including accuracy, recall and F1 value. Determine the importance of accuracy, recall, and F1 score based on the specific business needs of anomaly detection for corporate credit data. If the risk of missing anomaly companies is high, prioritize recall. If the cost of misclassifying normal companies as anomalies is high, prioritize accuracy. Use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner; Establish a feedback mechanism to compare the test results with the set data to optimize the model, and update the data and model regularly.

2. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 1, characterized in that: The collection of enterprise credit data and pre-processing of the data include: Collect corporate credit data and perform data cleaning on it; this data cleaning includes processing missing values, removing duplicate data and outliers; Select key features of corporate credit data and perform normalization and standardization.

3. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 1, characterized in that: The adaptive selection of anomaly detection algorithms based on the enterprise credit data type includes: Analyze corporate credit data types; According to the enterprise credit data type, the anomaly detection algorithm is adaptively selected according to the algorithm selection strategy.

4. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 1, characterized in that: The training and validation of the machine learning model includes: Get the training data and divide the dataset into training set and test set.

5. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 2, characterized in that: The collection of enterprise credit data and data cleaning of the data include: Conduct a comprehensive check on the enterprise credit data to identify the fields with missing values ​​and record the number of missing values ​​and the fields in which they are located; Analyze the reasons for missing values ​​and classify missing values ​​into missing types according to the reasons for missing values. Missing types include: random missing, completely random missing, and non-random missing. Select missing value processing method based on the proportion of missing values; Use set functions to identify duplicate data in corporate credit data and record the number and specific content of duplicate data; Delete or merge duplicate data; Identify outliers in corporate credit data and analyze the causes of outliers; Anomaly handling methods are used to handle outliers, including deleting outliers, replacing outliers, and binning. The selection of key features of enterprise credit data and normalization and standardization processing include: Select key features based on feature selection methods, including filtering methods, wrapper methods, and embedded methods; For each selected key feature, determine its normalization method; Calculate statistics of key features; Transform the eigenvalues ​​according to the selected normalization method; Select a standardization method; Calculate the mean and standard deviation of each key feature; For each eigenvalue, subtract the mean of the key feature from it, and then divide it by the standard deviation of the key feature to obtain the standardized eigenvalue.

6. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 3, characterized in that: The method of adaptively selecting an anomaly detection algorithm based on the enterprise credit data type and the algorithm selection strategy includes: For numerical data, choose statistical methods or machine learning methods; For data with time series characteristics, choose time series analysis methods or deep learning methods; For public opinion data, choose text processing and machine learning methods; The algorithm selection strategy includes: Perform feature analysis on the input corporate credit data to determine the data type, distribution characteristics, and dimensions, and calculate the statistical characteristics of the data; Build a library of anomaly detection algorithms targeting different data types; Encapsulate each algorithm to give it a unified interface; Select evaluation metrics to measure the performance of the algorithm, including accuracy, recall, and F1 value; For different types of data, choose different evaluation indicators; According to the results of feature analysis, select an algorithm suitable for the current data from the algorithm library; Use a portion of labeled data sets to evaluate the algorithm and select the algorithm based on the evaluation indicators; If the data type changes or new data is added, re-analysis of features and selection of algorithms are performed to make adaptive adjustments.

7. The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 4, characterized in that: For each split, the model is trained using the training set and the performance indicators of the model are evaluated on the validation set. The following steps are also included: Record the results of each validation and calculate the mean and standard deviation to evaluate model performance; For a specific anomaly detection model, determine the hyperparameters that need to be tuned; Combine the value range of each hyperparameter into a parameter grid; Use cross-validation to train and validate the model on each hyperparameter combination in the parameter grid; Record the verification results of each hyperparameter combination and select the hyperparameter combination; If the results obtained by the grid search do not meet the set requirements, further search for the hyperparameter combination is performed.

8. An artificial intelligence-based enterprise credit data anomaly detection device, characterized in that: The method for detecting anomalies in enterprise credit data based on artificial intelligence according to claim 1 is implemented, comprising: Data collection module, used to collect enterprise credit data and pre-process the data; Algorithm selection module, used to adaptively select anomaly detection algorithm based on the enterprise credit data type; Model training module, used to train and validate machine learning models; The anomaly detection module is used to use the trained model to perform real-time detection on newly input credit data, identify anomalies, and present the detection results in a visual manner; The feedback adjustment module is used to establish a feedback mechanism, compare the test results with the set data to optimize the model, and regularly update the data and model.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the enterprise credit data anomaly detection method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a server, implements the enterprise credit data anomaly detection method based on artificial intelligence as described in any one of claims 1 to 7.

Images