Abnormal alarm method, device, equipment and storage medium for business data
Through feature extraction and fusion analysis of business data, combined with information processing of MSCA mechanism, the problem of inaccurate abnormal identification in traditional methods is solved, the accuracy and efficiency of abnormal alarms are improved, and dynamic monitoring and timely monitoring of business data is achieved. response.
Patent Information
- Application Number
- CN202410312649.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-03-19
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-03-19
AI Technical Summary
When traditional exception alarm methods deal with high-dimensional and complex business data, they cannot accurately identify abnormalities, resulting in high false alarms or missed response rates, which cannot meet the requirements of modern business systems for abnormal detection.
By collecting business data, extracting feature information, performing fusion analysis, generating strategy information, and using the preset MSCA mechanism to perform information analysis to determine whether the alarm conditions are met. If it is met, the alarm processing will be carried out, including multi-scale convolutional blocks and SE modules , Softmax function and residual network and other technical means.
It improves the accuracy and efficiency of abnormal alarms, reduces the false alarm rate and missed alarm rate, and realizes the dynamic and real-time monitoring of business data abnormalities in time to respond to business abnormal events in a timely manner.
Smart Images

Figure CN118113566B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of business data technology, and in particular to a business data abnormality alarm method, device, equipment and storage medium. Background Art
[0002] With the rapid development of information technology, business systems are becoming more and more complex. In a changing business environment, various abnormal events occur frequently, which brings great challenges to the operation and management of enterprises. Especially in the fields of finance, communications, network security, etc., once a business abnormality occurs, it may cause huge economic losses or even serious damage to brand reputation. Therefore, timely detection and processing of abnormalities in business data has become an important means to improve business stability and ensure user experience.
[0003] However, traditional anomaly alarm methods often rely on simple threshold judgments or fixed pattern matching, lacking in-depth understanding of the complexity of business data and flexible responses. When processing high-dimensional, complex-pattern business data, these methods often fail to accurately identify anomalies, resulting in high false positive or false negative rates, and fail to meet the requirements of modern business systems for anomaly detection. Summary of the invention
[0004] The main purpose of the present invention is to provide a method, device, equipment and storage medium for abnormal alarm of business data, so as to improve the accuracy and efficiency of abnormal alarm and reduce the false alarm rate.
[0005] To achieve the above object, the present invention provides a method for abnormal alarm of business data, comprising the following steps:
[0006] Collect business data sent by business systems;
[0007] Extracting characteristic information from the business data to obtain business result characteristic information;
[0008] Performing a fusion analysis on the business data and the business result feature information to obtain an analysis result of the business data, and generating strategy information based on the analysis result of the business data;
[0009] If the policy information meets the alarm condition, the policy information is used to perform alarm processing; wherein the alarm processing at least includes an alarm method.
[0010] As a further solution of the present invention, feature information is extracted from the service data to obtain service result feature information, including:
[0011] Extracting the business data through a preset neural network to obtain business feature data;
[0012] Performing weight analysis on the service characteristic data to obtain a characteristic level of the service data;
[0013] Determining multiple criticality information of the business data based on the characteristic level of the business data;
[0014] Feature weighting is performed on multiple criticality information of the business data to obtain business result feature information.
[0015] As a further solution of the present invention, the business data and the business result feature information are fused and analyzed to obtain analysis results, including:
[0016] Fusion analysis is performed on the business data and the business result feature information to obtain time series features of the business data;
[0017] Constructing a time series feature label matrix according to the time series features of the business data, performing cluster center calculation on the time series feature label matrix, and obtaining a variance value corresponding to the time series feature label matrix;
[0018] Based on the variance value corresponding to the time series feature label matrix and a preset standard variance value, the time series feature label matrix is analyzed to obtain an analysis result corresponding to the time series feature label matrix.
[0019] As a further solution of the present invention, the generating of the policy information based on the analysis result of the business data includes:
[0020] Inputting the analysis results of the business data into a preset mining model for data mining to obtain preliminary mining results; wherein the preset mining model is a pre-trained model; the preliminary mining results include a first preliminary mining result and a second preliminary mining result, the first preliminary mining result has corresponding policy information, and the second preliminary mining result does not have corresponding policy information;
[0021] Inputting the business data into a preset deep mining model to perform data deep mining to obtain second deep mining business data;
[0022] Performing data clustering analysis on the second deep mining business data to obtain types of the business data;
[0023] Calculating the degree of abnormality of the business data corresponding to the second preliminary mining result to obtain an abnormal result of the business data;
[0024] Performing time window analysis on the business data corresponding to the second preliminary mining result to obtain business data for each time window;
[0025] Performing fusion analysis on the types of the business data, abnormal results of the business data, and business data of each time window to obtain fusion analysis results of the business data;
[0026] Inputting the fusion analysis result of the business data into a preset mining model to perform data mining to obtain a third preliminary mining result;
[0027] It is determined whether the third preliminary mining result has corresponding strategy information. If the third preliminary mining result does not have corresponding strategy information, the third preliminary mining result is input into a preset deep mining model to continue data deep mining until the corresponding strategy information is generated.
[0028] As a further solution of the present invention, if the alarm condition is met in the policy information, the policy information is used for processing, including:
[0029] Performing information analysis on the policy information through a preset MSCA mechanism to obtain a policy analysis result; wherein the preset MSCA mechanism is a pre-trained mechanism;
[0030] Comparing the strategy analysis results with preset alarm conditions;
[0031] If the policy analysis result meets the alarm condition, the policy information is used for processing.
[0032] As a further solution of the present invention, the preset MSCA mechanism includes a multi-scale convolution block, a SE module, a Softmax function and a residual network, the output end of the multi-scale convolution block is connected to the input end of the SE module, the output end of the SE module is connected to the input end of the Softmax function, the output end of the Softmax function is connected to the input end of the residual network, the residual network includes three layers of CNN layers and three layers of ResNet, the output ends of the three layers of CNN layers are all connected to the output ends of the Softmax function, and the input ends of the three layers of ResNet are all correspondingly connected to the output ends of the three layers of CNN layers; the information analysis of the policy information input by the preset MSCA mechanism to obtain the policy analysis result includes:
[0033] Extracting the strategy information through a multi-scale convolution block to obtain strategy multi-scale information;
[0034] The SE module is used to associate the multi-scale information of the strategy with the feature channel dimension to obtain the attention vector of the feature channel dimension;
[0035] Recalibrate the attention vector of the feature channel dimension by using a Softmax function to obtain a calibration weight of the feature channel dimension;
[0036] Multiplying the calibration weight of the feature channel dimension by the strategy multi-scale information to obtain weighted strategy multi-scale information;
[0037] The weighted multi-scale information of the strategy is input into the three-layer CNN layer in the residual network to capture the information receptive field, and the feature information of multiple receptive field dimensions is obtained;
[0038] The feature information of the multiple receptive field dimensions is input into a preset three-layer ResNet for nonlinear transformation to obtain a strategy analysis result.
[0039] The present invention also provides a device for abnormal alarm of business data, comprising:
[0040] A collection module is used to collect business data sent by the business system;
[0041] An extraction module, used to extract feature information from the business data to obtain business result feature information;
[0042] An analysis module, used to perform a fusion analysis on the service data and the service result characteristic information to obtain an analysis result of the service data, and generate strategy information based on the analysis result of the service data;
[0043] The judgment module is used to use the policy information to perform alarm processing if the policy information meets the alarm condition; wherein the alarm processing at least includes the alarm method.
[0044] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.
[0045] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.
[0046] The present invention provides an abnormal alarm method, device, equipment and storage medium for business data, comprising the following steps: collecting business data sent by a business system; extracting feature information from the business data to obtain business result feature information; fusing and analyzing the business data and the business result feature information to obtain analysis results of the business data, and generating policy information based on the analysis results of the business data; if the policy information meets the alarm condition, the policy information is used for alarm processing; wherein the alarm processing at least includes an alarm method; by performing in-depth analysis on the collected business data, key feature information that can represent the business operation status is extracted, thereby solving the technical problem of being unable to accurately identify anomalies, resulting in high false alarm or missed alarm rates, improving the accuracy and efficiency of abnormal alarms, reducing false alarm rates and missed alarm rates, and achieving dynamic and real-time monitoring of business data anomalies, and achieving the beneficial effect of timely responding to business abnormal events. BRIEF DESCRIPTION OF THE DRAWINGS
[0047] Figure 1 It is a schematic diagram of the steps of a method for abnormal alarm of business data in one embodiment of the present invention;
[0048] Figure 2 It is a structural block diagram of an abnormal alarm device for business data in one embodiment of the present invention;
[0049] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.
[0050] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION
[0051] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.
[0052] like Figure 1 As shown, Figure 1 This is a schematic diagram of the steps of a method for abnormal alarm of business data in one embodiment of the present invention;
[0053] An embodiment of the present invention provides a method for abnormal alarm of business data, comprising the following steps:
[0054] Step S1, collecting business data sent by the business system.
[0055] Specifically, determine the business system or component that will serve as the data source, identify and define the data type and format that needs to be collected, such as transaction records, user behavior, system logs, etc. Establish a data interface or communication link with the business system so that data can flow smoothly into the data collection system. The above stage involves technologies such as application programming interface (API), message queue, log forwarding, etc. Use data crawling tools or services to extract data from the business system according to predetermined rules and schedules. The above data can be real-time, batch, or event-driven. Preprocess the collected raw data, including removing irrelevant information, handling missing values, formatting adjustments, etc., to ensure data quality and facilitate subsequent analysis. Save the cleaned data to a data warehouse or database for easy query and analysis, and select a suitable storage solution, such as a relational database, NoSQL database, data lake, etc. Implement permission management and data access control to ensure data security. Implement data backup and recovery strategies to prevent data loss. Monitor the data collection process, discover and handle abnormal events in a timely manner, and continuously optimize data collection processes and tools based on system performance and needs.
[0056] The above steps can achieve the following technical effects: Through a systematic data collection process, it can ensure that the collected business data is both comprehensive and accurate, providing a solid foundation for subsequent analysis. Real-time or near-real-time data capture can help business systems quickly respond to market and operational changes and improve decision-making efficiency. Data cleaning and standardization can significantly improve data quality and reduce analysis errors caused by data problems. Well-designed data collection processes can be easily expanded to more data sources to meet the needs of business development and changes. Manage data access and backup to ensure data security and compliance with relevant laws and regulations. Provide timely and accurate data support for enterprises, enhance the ability of data-driven decision-making, and thus affect the strategic planning and daily operations of the entire enterprise.
[0057] Step S2: collecting business data sent by the business system and extracting feature information from the business data to obtain business result feature information.
[0058] Specifically, the collected business data is cleaned, including removing outliers, filling missing values, and normalizing data to improve data quality, and the data is formatted to ensure that it meets the input requirements of the feature extraction algorithm. Determine which data attributes are critical to understanding business results. Include statistical analysis of the data to identify features with high relevance and predictive value. Use techniques such as principal component analysis (PCA) and automatic feature selection algorithms to reduce data dimensions and focus on the most important features. Based on business understanding and data analysis, new features are constructed, including combining existing data fields and generating derived variables. Use domain knowledge to transform business logic into quantifiable data features. Apply machine learning algorithms (such as decision trees, neural networks, cluster analysis, etc.) to automatically extract and identify key patterns and features in the data. Perform time series analysis, text analysis, etc. to extract relevant features based on data types. Evaluate the relevance and predictive power of features to business results, and optimize features through feedback from machine learning models. Adjust and optimize the feature extraction process to ensure the effectiveness and accuracy of features. Integrate the extracted features into the business analysis model for testing and verification. Verify the effectiveness and prediction accuracy of feature information through actual business scenarios.
[0059] The above steps can achieve the following technical effects: Through in-depth analysis and feature extraction, we can better understand the patterns and trends behind business data. Accurate feature extraction helps to build more accurate prediction models and effectively predict future business results. By focusing on key features, we can reduce the interference of irrelevant data noise and improve the performance of the model. It can handle diverse and dynamically changing business data and adapt to changing business needs. It provides decision makers with more accurate and insightful data analysis results to support more effective business decisions. Through accurate business forecasting and trend analysis, it helps enterprises allocate resources and plan strategies more effectively. In summary: Through efficient data processing and intelligent analysis, enterprises can enhance their understanding and utilization of business data, providing strong support for more intelligent and accurate business operations.
[0060] Step S3: perform a fusion analysis on the business data and the business result feature information to obtain an analysis result of the business data, and generate policy information based on the analysis result of the business data.
[0061] Specifically, collect relevant business data and business result data. The above data may include sales records, customer feedback, market trends, etc. Organize data to ensure data quality, including cleaning data, processing missing values and outliers. Extract features from business data, which should effectively reflect business conditions and business results. Fusion of business data features and business result features, including integration of features from different data sources to form a comprehensive feature set. Conduct exploratory data analysis (EDA) to discover potential relationships and patterns between data. Build a prediction or classification model based on the fused data. The above model may be a machine learning model, such as decision tree, random forest, neural network, etc. Use cross-validation and other methods to evaluate model performance to ensure the accuracy and generalization ability of the model. Adjust and optimize the model according to the evaluation results. Generate business strategy information based on the model analysis results. Business strategy information may be specific suggestions on product optimization, market positioning, customer relationship management, etc. Concrete implementation of the strategy, such as presentation to decision makers through reports, dashboards, etc. After implementing the strategy, continuously monitor business performance and collect new business data. Iterate and optimize models and strategies based on new business data and business performance.
[0062] The above steps can achieve the following technical effects: Generate strategies through data-driven methods, and make evidence-based decisions faster. Fusion analysis can reveal the deep patterns and relationships behind business data, providing enterprises with deeper insights. Strategies based on analysis results can help enterprises allocate resources more effectively, such as investing in markets or products with greater potential. Risk reduction: Through prediction and classification models, enterprises can foresee potential risks and challenges, so as to prepare in advance and reduce losses. By analyzing customer data and feedback, enterprises can better understand customer needs and preferences, thereby providing more personalized services or products, and increasing customer satisfaction and loyalty. Through such steps and methods, enterprises can more systematically analyze and understand complex business environments, formulate more precise and effective business strategies, and thus gain advantages in the highly competitive market environment.
[0063] Step S4: if the policy information meets the alarm condition, the policy information is used to perform alarm processing; wherein the alarm processing at least includes an alarm method.
[0064] Specifically, alarm thresholds are set based on business needs and historical data analysis. Alarm thresholds can be upper or lower limits of business indicators, and alarms are triggered when these ranges are exceeded. When the results of policy analysis exceed the preset alarm thresholds, the system automatically generates an alarm signal. The alarm signal can clearly indicate the specific business indicator and reason that triggered the alarm. The alarm information is notified to relevant decision makers or operation teams through predetermined channels. Notification channels can include email, SMS, in-app notifications, etc. The team or individual receiving the alarm needs to confirm the alarm, analyze the cause of the alarm, and take corresponding response measures based on the urgency and severity of the alarm. Solve the problem that triggered the alarm, and record the process and results of the problem solving in detail for future reference and improvement. According to the effectiveness and accuracy of alarm processing, iteratively optimize the alarm conditions, notification methods, and response processes to improve the efficiency and accuracy of the alarm system.
[0065] The above steps can achieve the following technical effects: Through real-time monitoring and alarms, potential business risks can be quickly identified and responded to to prevent problems from expanding. Timely alarm and response mechanisms help maintain the continuity of business operations and reduce the negative impact of emergencies on the business. The real-time data analysis and feedback provided by the alarm system can assist decision makers in making more timely and accurate decisions. Through the analysis of alarm data, enterprises can allocate and optimize resources more effectively, such as concentrating resources where they are more needed to avoid waste. The automation of alarm processing not only improves efficiency, but also reduces human errors and improves the intelligence level of the entire system. Through such alarm processing steps, enterprises can build a more sensitive and effective business monitoring and response mechanism to ensure the stability and security of business operations.
[0066] In a specific implementation manner, feature information is extracted from the business data to obtain business result feature information, including:
[0067] Extracting the business data through a preset neural network to obtain business feature data;
[0068] Performing weight analysis on the service characteristic data to obtain a characteristic level of the service data;
[0069] Determining multiple criticality information of the business data based on the characteristic level of the business data;
[0070] Feature weighting is performed on multiple criticality information of the business data to obtain business result feature information.
[0071] Specifically, before extracting features, business data usually needs to be preprocessed, including data cleaning (removing noise and outliers), standardization, normalization, etc., to ensure data quality and consistency. Use a preset neural network model (such as convolutional neural network (CNN) for image data and recurrent neural network (RNN) for time series data) to automatically extract data features. The above steps rely on deep learning technology to automatically learn high-level features of data through multiple layers of the network. Perform weight analysis on the feature data extracted by the neural network to determine the degree of influence of each feature on the business results. This is achieved through feature importance evaluation technology, such as model-based feature importance scoring (for example, feature importance in random forests). Based on the results of the weight analysis, the features are graded to determine the criticality of different features on the business results. The above steps can be achieved by setting thresholds or using clustering algorithms to identify the most influential feature subsets. Weight the key features, assign different weights according to their degree of influence on the business results, and fuse these weighted features into comprehensive business result feature information. The above can be achieved through simple weighted average or more complex mathematical models to obtain the final business result feature representation.
[0072] The above steps can achieve the following technical effects: By accurately extracting and weighting key features, the accuracy and efficiency of subsequent model training can be improved, because the model can focus on learning the factors that have the greatest impact on the results. By identifying and analyzing key features, the interpretability of the model can be enhanced, helping business decision makers understand the decision basis of the model. Clarifying key features can help companies allocate resources more effectively, for example, allocating more marketing budgets to channels that have the greatest impact on sales results. Through automated feature extraction and analysis processes, the transformation process from data to decisions can be accelerated, improving business agility and responsiveness. By monitoring and analyzing key features, companies can better predict and manage potential business risks. The implementation of this process helps companies deeply understand the inherent structure and dynamics of their business data, so as to make more informed decisions in an increasingly competitive market environment.
[0073] In a specific implementation manner, the business data and the business result feature information are fused and analyzed to obtain analysis results, including:
[0074] Fusion analysis is performed on the business data and the business result feature information to obtain time series features of the business data;
[0075] Constructing a time series feature label matrix according to the time series features of the business data, performing cluster center calculation on the time series feature label matrix, and obtaining a variance value corresponding to the time series feature label matrix;
[0076] Based on the variance value corresponding to the time series feature label matrix and a preset standard variance value, the time series feature label matrix is analyzed to obtain an analysis result corresponding to the time series feature label matrix.
[0077] Specifically, first, the original business data and the business result feature information extracted by advanced analysis techniques (such as machine learning, deep learning, etc.) are comprehensively analyzed. Special attention is paid to using the time series nature of business data to explore the key patterns and trends of data changes over time. Based on the time series characteristics of business data, a time series feature label matrix is constructed. The time series feature label matrix organizes the time series features in the form of labels to facilitate the subsequent analysis process. Cluster analysis is performed on the time series feature label matrix, the center of each cluster is calculated, and the corresponding variance value is obtained accordingly. The above variance value reflects the degree of feature difference between different categories, that is, the distribution divergence of the data. The obtained variance value is compared with the preset standard variance value. Based on this comparison result, the time series feature label matrix is deeply analyzed to obtain the final analysis result. The purpose of this step is to identify features that deviate significantly from the normal range, so as to determine possible anomalies.
[0078] The above steps can achieve the following technical effects: By utilizing the time series characteristics of business data, the anomaly detection model can more sensitively capture abnormal behaviors or trends that change over time, thereby improving the timeliness and accuracy of anomaly warnings. By constructing a time series feature label matrix and performing cluster analysis and variance comparison, the data analysis process can be dynamically adjusted according to the characteristics of the data itself, thereby improving the accuracy and reliability of the analysis. By calculating the variance value and comparing it with the preset standard value, the difference in data characteristics can be quantified. The above quantitative difference analysis helps to identify and determine abnormal situations more objectively and accurately. The final analysis results help to reveal the deep problems and laws behind the business data, provide data-driven decision support for enterprises, and thus play an important role in risk warning, business optimization, etc. In summary, by conducting in-depth fusion analysis of business data and business result feature information, the effectiveness, timeliness and accuracy of abnormal alarms can be greatly improved, providing strong technical support for the continuous operation of the business.
[0079] In a specific implementation, the generating of policy information based on the analysis result of the business data includes:
[0080] Inputting the analysis results of the business data into a preset mining model for data mining to obtain preliminary mining results; wherein the preset mining model is a pre-trained model; the preliminary mining results include a first preliminary mining result and a second preliminary mining result, the first preliminary mining result has corresponding policy information, and the second preliminary mining result does not have corresponding policy information;
[0081] Inputting the business data into a preset deep mining model to perform data deep mining to obtain second deep mining business data;
[0082] Performing data clustering analysis on the second deep mining business data to obtain types of the business data;
[0083] Calculating the degree of abnormality of the business data corresponding to the second preliminary mining result to obtain an abnormal result of the business data;
[0084] Performing time window analysis on the business data corresponding to the second preliminary mining result to obtain business data for each time window;
[0085] Performing fusion analysis on the types of the business data, abnormal results of the business data, and business data of each time window to obtain fusion analysis results of the business data;
[0086] Inputting the fusion analysis result of the business data into a preset mining model to perform data mining to obtain a third preliminary mining result;
[0087] It is determined whether the third preliminary mining result has corresponding strategy information. If the third preliminary mining result does not have corresponding strategy information, the third preliminary mining result is input into a preset deep mining model to continue data deep mining until the corresponding strategy information is generated.
[0088] Specifically, first, the analysis results of the business data are input into a pre-trained mining model (such as decision tree, random forest, neural network, etc.) to perform preliminary data mining to obtain preliminary mining results. The preliminary mining results are divided into two categories: the first category has corresponding policy information, while the second category does not. For those results that cannot be directly associated with policy information in the preliminary mining stage (the second preliminary mining results), a more complex or more specialized deep mining model is used for further analysis to obtain deeply mined business data. Cluster analysis is performed on some of the deeply mined business data to identify data types and calculate their abnormality. In addition, time window analysis is performed on the data to explore business dynamics in different time periods. The data type, abnormal results and time window analysis results are fused and analyzed to obtain a comprehensive analysis result. The fusion analysis results are input into the mining model again for data mining. If the corresponding policy information cannot be generated at this time, the deep mining model is continued to be used for further deep mining until effective policy information can be generated.
[0089] The above steps can achieve the following technical effects: By combining preliminary mining with deep mining, the method can adopt different depth mining strategies for different business data, thereby enhancing the depth and breadth of analysis and ensuring that more potential business models are identified and utilized. Through deeper data analysis, more accurate and detailed strategy information can be generated to provide reliable data support for business decisions and improve the accuracy and reliability of decisions. Through time window analysis and abnormality degree calculation, the method can more sensitively capture changes in business trends and the emergence of abnormal patterns, providing a basis for timely adjustment of business strategies. The feedback mechanism of the process ensures that data that has not yet obtained strategy information after preliminary mining is not omitted, and through continuous deep mining until the corresponding strategy is found, dynamic and iterative strategy adjustment and optimization are achieved. In summary, by combining preliminary mining with deep mining, the method not only enhances the depth and breadth of business data analysis, but also improves the accuracy of business decisions and the dynamic adjustment ability of strategies, and has important application value for handling complex and changing business scenarios.
[0090] In a specific implementation, if the policy information satisfies the alarm condition, the policy information is used for processing, including:
[0091] Performing information analysis on the policy information through a preset MSCA mechanism to obtain a policy analysis result; wherein the preset MSCA mechanism is a pre-trained mechanism;
[0092] Comparing the strategy analysis results with preset alarm conditions;
[0093] If the policy analysis result meets the alarm condition, the policy information is used for processing.
[0094] Specifically, the policy information is input into the pre-trained MSCA mechanism. The MSCA mechanism is designed to conduct in-depth and comprehensive analysis of policy information. The mechanism extracts key policy analysis results by integrating multi-source information and a deep understanding of the current business context. Subsequently, the policy analysis results obtained by the MSCA mechanism are compared with the pre-set alarm conditions. These alarm conditions may include but are not limited to data thresholds, abnormal patterns, risk levels, and other indicators related to business operation security and efficiency. If the policy analysis results meet the preset alarm conditions, the system will take predetermined response measures, including sending alarms, initiating abnormal handling processes, adjusting operating parameters, etc., to deal with the identified potential risks or abnormal conditions.
[0095] The above steps can achieve the following technical effects: through the MSCA mechanism combined with multi-source information fusion and situational awareness, the depth and accuracy of policy information analysis are greatly improved, and the key dynamics in complex business environments can be captured more effectively. Compared with the traditional alarm system based on simple rules, the alarm mechanism of this embodiment is more intelligent and sensitive, and can trigger alarms based on complex analysis results, and discover and respond to potential risks and anomalies in advance. This processing flow enables the business system to respond quickly and take appropriate countermeasures when facing risks or anomalies corresponding to specific policy information, thereby improving the reliability and robustness of the business. By continuously analyzing policy information and feedback results, as well as comparative analysis with alarm conditions, business operation strategies can be dynamically adjusted and optimized according to real-time business scenarios and risk assessments. In summary, by introducing advanced MSCA mechanisms and intelligent alarm decision logic, the business system is given better situational awareness, risk identification and processing capabilities, providing strong support for maintaining the security and stability of business operations.
[0096] In a specific implementation, the preset MSCA mechanism includes a multi-scale convolution block, a SE module, a Softmax function and a residual network, the output end of the multi-scale convolution block is connected to the input end of the SE module, the output end of the SE module is connected to the input end of the Softmax function, the output end of the Softmax function is connected to the input end of the residual network, the residual network includes three layers of CNN layers and three layers of ResNet, the output ends of the three layers of CNN layers are all connected to the output ends of the Softmax function, and the input ends of the three layers of ResNet are all correspondingly connected to the output ends of the three layers of CNN layers; the information analysis of the policy information input by the preset MSCA mechanism to obtain the policy analysis result includes:
[0097] Extracting the strategy information through a multi-scale convolution block to obtain strategy multi-scale information;
[0098] The SE module is used to associate the multi-scale information of the strategy with the feature channel dimension to obtain the attention vector of the feature channel dimension;
[0099] Recalibrate the attention vector of the feature channel dimension by using a Softmax function to obtain a calibration weight of the feature channel dimension;
[0100] Multiplying the calibration weight of the feature channel dimension by the strategy multi-scale information to obtain weighted strategy multi-scale information;
[0101] The weighted multi-scale information of the strategy is input into the three-layer CNN layer in the residual network to capture the information receptive field, and the feature information of multiple receptive field dimensions is obtained;
[0102] The feature information of the multiple receptive field dimensions is input into a preset three-layer ResNet for nonlinear transformation to obtain a strategy analysis result.
[0103] Specifically, by processing policy information through multi-scale convolution blocks, the representation of policy features at different scales can be captured, improving the system's ability to capture policy information details and context. Multi-scale convolution blocks allow the model to analyze policy information within receptive fields of different sizes, enhancing the model's feature extraction capabilities. The SE module focuses on learning the relationship between different feature channels. By generating attention vectors for feature channels, important features can be strengthened and irrelevant features can be suppressed, thereby significantly improving the model's expressiveness without increasing too much computational complexity. Using the Softmax function to calibrate the weights of feature channels can effectively redistribute the importance of each channel, allowing the model to focus more on information that is beneficial to policy analysis. Through the application of residual networks, especially in combination with multi-layer CNN and ResNet structures, the problem of information degradation in deep networks can be effectively handled, the fluidity of information in the deep layers of the network can be enhanced, and the problem of gradient disappearance or explosion can be avoided, enabling the model to capture deeper and more complex policy information.
[0104] The above steps can achieve the following technical effects: Through the combination of multi-scale convolution and SE modules, the model can better capture and represent policy features at different levels and scales, and improve the representation quality of policy information. The use of Softmax function and residual network structure enables the model to have better accuracy and stability when processing complex policy information, while improving the interpretability of the model. Through the optimized design of SE module and residual network, the model can achieve in-depth analysis of policy information while maintaining low computational complexity, effectively improving the efficiency and accuracy of information analysis. In summary, through this advanced MSCA mechanism, in-depth and comprehensive analysis of policy information can be achieved, providing more accurate and efficient decision-making basis for decision support system, and further enhancing the system's adaptability to complex environments and information processing capabilities.
[0105] The above describes the abnormal alarm method for service data in the embodiment of the present invention. The following describes the abnormal alarm device for service data in the embodiment of the present invention. Figure 2 , an embodiment of the abnormal alarm device for business data in an embodiment of the present invention includes:
[0106] The acquisition module 21 is used to acquire the business data sent by the business system;
[0107] An extraction module 22 is used to extract feature information from the business data to obtain business result feature information;
[0108] An analysis module 23 is used to perform a fusion analysis on the service data and the service result characteristic information to obtain an analysis result of the service data, and generate policy information based on the analysis result of the service data;
[0109] The judgment module 24 is used to use the policy information to perform alarm processing if the policy information meets the alarm condition; wherein the alarm processing at least includes the alarm mode.
[0110] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.
[0111] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.
[0112] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.
[0113] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.
[0114] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.
[0115] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.
[0116] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for abnormal alarm of business data, characterized in that: The following steps are involved: Collect business data sent by business systems; Extracting characteristic information from the business data to obtain business result characteristic information; Performing a fusion analysis on the business data and the business result feature information to obtain an analysis result of the business data, and generating strategy information based on the analysis result of the business data; If the policy information meets the alarm condition, the policy information is used to perform alarm processing; wherein the alarm processing at least includes an alarm method; Extracting characteristic information from the business data to obtain business result characteristic information includes: Extracting the business data through a preset neural network to obtain business feature data; Performing weight analysis on the service characteristic data to obtain a characteristic level of the service data; Determining multiple criticality information of the business data based on the characteristic level of the business data; Performing feature weighting on multiple criticality information of the business data to obtain business result feature information; The business data and the business result feature information are integrated and analyzed to obtain analysis results of the business data, including: Fusion analysis is performed on the business data and the business result feature information to obtain time series features of the business data; Constructing a time series feature label matrix according to the time series features of the business data, performing cluster center calculation on the time series feature label matrix, and obtaining a variance value corresponding to the time series feature label matrix; Based on the variance value corresponding to the time series feature label matrix and the preset standard variance value, the time series feature label matrix is analyzed to obtain the analysis result of the business data corresponding to the time series feature label matrix.
2. The abnormal alarm method for business data according to claim 1 is characterized in that: The generating of the policy information based on the analysis result of the business data includes: Inputting the analysis results of the business data into a preset mining model for data mining to obtain preliminary mining results; wherein the preset mining model is a pre-trained model; the preliminary mining results include a first preliminary mining result and a second preliminary mining result, the first preliminary mining result has corresponding policy information, and the second preliminary mining result does not have corresponding policy information; Inputting the business data into a preset deep mining model to perform data deep mining to obtain second deep mining business data; Performing data clustering analysis on the second deep mining business data to obtain types of the business data; Calculating the degree of abnormality of the business data corresponding to the second preliminary mining result to obtain an abnormal result of the business data; Performing time window analysis on the business data corresponding to the second preliminary mining result to obtain business data for each time window; Performing fusion analysis on the types of the business data, abnormal results of the business data, and business data of each time window to obtain fusion analysis results of the business data; Inputting the fusion analysis result of the business data into a preset mining model to perform data mining to obtain a third preliminary mining result; It is determined whether the third preliminary mining result has corresponding strategy information. If the third preliminary mining result does not have corresponding strategy information, the third preliminary mining result is input into a preset deep mining model to continue data deep mining until the corresponding strategy information is generated.
3. The abnormal alarm method for business data according to claim 1 is characterized in that: If the policy information meets the alarm conditions, the policy information is used for alarm processing, including: The policy information is analyzed by a preset MSCA mechanism to obtain a policy analysis result; wherein the preset MSCA mechanism is a pre-trained mechanism; the preset MSCA mechanism includes a multi-scale convolution block, an SE module, a Softmax function and a residual network, the output end of the multi-scale convolution block is connected to the input end of the SE module, the output end of the SE module is connected to the input end of the Softmax function, the output end of the Softmax function is connected to the input end of the residual network, the residual network includes three CNN layers and three ResNet layers, the output ends of the three CNN layers are all connected to the input end of the Softmax function, and the input ends of the three ResNet layers are all correspondingly connected to the output ends of the three CNN layers; Comparing the strategy analysis results with preset alarm conditions; If the policy analysis result meets the alarm condition, the policy information is used for alarm processing.
4. The abnormal alarm method for business data according to claim 3 is characterized in that: The performing information analysis on the policy information by using a preset MSCA mechanism to obtain a policy analysis result includes: Extracting the strategy information through a multi-scale convolution block to obtain strategy multi-scale information; The SE module is used to associate the multi-scale information of the strategy with the feature channel dimension to obtain the attention vector of the feature channel dimension; Recalibrate the attention vector of the feature channel dimension through the Softmax function to obtain the calibration weight of the feature channel dimension; Multiplying the calibration weight of the feature channel dimension by the strategy multi-scale information to obtain weighted strategy multi-scale information; The weighted multi-scale information of the strategy is input into the three-layer CNN layer in the residual network to capture the information receptive field, and the feature information of multiple receptive field dimensions is obtained; The feature information of the multiple receptive field dimensions is input into a preset three-layer ResNet for nonlinear transformation to obtain a strategy analysis result.
5. A device for abnormal alarm of business data, characterized in that: include: A collection module is used to collect business data sent by the business system; An extraction module, used to extract feature information from the business data to obtain business result feature information; An analysis module, used to perform a fusion analysis on the service data and the service result characteristic information to obtain an analysis result of the service data, and generate strategy information based on the analysis result of the service data; A judgment module, used for using the strategy information to perform alarm processing if the strategy information meets the alarm condition; wherein the alarm processing at least includes an alarm mode; Extracting characteristic information from the business data to obtain business result characteristic information includes: Extracting the business data through a preset neural network to obtain business feature data; Performing weight analysis on the service characteristic data to obtain a characteristic level of the service data; Determining multiple criticality information of the business data based on the characteristic level of the business data; Performing feature weighting on multiple criticality information of the business data to obtain business result feature information; The business data and the business result feature information are integrated and analyzed to obtain analysis results of the business data, including: Fusion analysis is performed on the business data and the business result feature information to obtain time series features of the business data; Constructing a time series feature label matrix according to the time series features of the business data, performing cluster center calculation on the time series feature label matrix, and obtaining a variance value corresponding to the time series feature label matrix; Based on the variance value corresponding to the time series feature label matrix and the preset standard variance value, the time series feature label matrix is analyzed to obtain the analysis result of the business data corresponding to the time series feature label matrix.
6. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 4 are implemented.
