Early warning service information generation method and system based on multi-dimensional service monitoring

Abstract

The invention relates to the technical field of data analysis, and provides an early warning service information generation method and system based on multi-dimensional service monitoring, through integration of multi-source data, the method provides a more comprehensive and multi-dimensional monitoring function, and effectively makes up for the defects of a traditional method. Meanwhile, by means of an advanced multi-dimensional knowledge mining algorithm, key information in the log data can be deeply analyzed and mined, and foreseeable early warning monitoring recognition viewpoints are generated. Therefore, not only is the real-time monitoring capability of an enterprise on the business state improved, but also the accuracy and efficiency of risk prevention and control are remarkably enhanced, and powerful technical support is provided for stable operation of the enterprise.

Description

Technical Field

[0001] The present application relates to the technical field of data analysis, and in particular to a method and system for generating early warning service information based on multi-dimensional business monitoring. Background Art

[0002] With the rapid development of online business, enterprises are facing increasingly complex operating environments and higher risk management requirements. Traditional business monitoring methods are often limited to a single data source or dimension, making it difficult to fully grasp the real-time operating status and potential risks of the business. In addition, the growing volume and complexity of data also pose higher challenges to log analysis. Simple keyword search or statistical analysis can no longer meet the in-depth needs of modern enterprises for business monitoring. Summary of the invention

[0003] In order to improve the above problems, the present application provides a method and system for generating early warning service information based on multi-dimensional business monitoring.

[0004] In a first aspect of an embodiment of the present application, a method for generating early warning service information based on multi-dimensional business monitoring is provided, which is applied to an early warning service information generation system. The method includes:

[0005] Obtain online business operation logs to be monitored for early warning;

[0006] By completing the debugging of the target multidimensional knowledge mining algorithm, multidimensional knowledge mining processing is performed on the online business operation log to be monitored for early warning, so as to obtain a target multidimensional knowledge vector;

[0007] According to the target multi-dimensional knowledge vector, an early warning monitoring identification viewpoint of the online business operation log to be subjected to early warning monitoring is obtained.

[0008] Optionally, the method further includes:

[0009] Using the multidimensional knowledge mining algorithm to be debugged, multidimensional knowledge mining processing is performed on the first online business operation log debugging sample to obtain a first business log multidimensional knowledge vector, wherein the first online business operation log debugging sample corresponds to a target business indicator keyword;

[0010] Determine a second business log multidimensional knowledge vector according to the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector, wherein the target attention embedded dense semantic vector is determined from the attention feature relationship set of the completed debugging, wherein the attention feature relationship set of the completed debugging includes at least one attention embedded dense semantic vector corresponding to each associated attention dimension, and each of the attention embedded dense semantic vectors respectively represents a monitoring warning description vector of the online business operation log corresponding to the target business indicator keyword in the corresponding associated attention dimension;

[0011] Using a multivariate decision tree algorithm, performing multivariate decision discrimination on the multidimensional knowledge vector of the second business log to obtain a multivariate decision discrimination label;

[0012] According to the multivariate decision discrimination label, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once to obtain the debugged multidimensional knowledge mining algorithm.

[0013] Optionally, optimizing the algorithm configuration variables of the multidimensional knowledge mining algorithm at least once based on the multivariate decision discrimination label to obtain the debugged multidimensional knowledge mining algorithm includes:

[0014] Determining a first target training error variable according to the multivariate decision discriminant label and the set prior training annotation data;

[0015] According to the first target training error variable, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once to obtain the debugged multidimensional knowledge mining algorithm.

[0016] Optionally, the set prior training annotation data includes the prior keyword annotation data of the target business indicator keyword and the first associated attention dimension prior training annotation data of the target attention embedded dense semantic vector, and the multivariate decision discrimination label includes business indicator keyword data and log associated attention dimension information; the first target training error variable is determined based on the multivariate decision discrimination label and the set prior training annotation data, including:

[0017] Determining a first training error variable according to the business indicator keyword data and the prior keyword labeling data;

[0018] Determining a second training error variable according to the log-associated attention dimension information and the first-associated attention dimension prior training annotation data;

[0019] The first target training error variable is determined according to the first training error variable and the second training error variable.

[0020] Optionally, the multidimensional knowledge mining algorithm includes a pyramid algorithm branch and an implicit algorithm branch; the multidimensional knowledge mining algorithm to be debugged is used to perform multidimensional knowledge mining processing on the first online business operation log debugging sample to obtain the first business log multidimensional knowledge vector, including:

[0021] Through the pyramid algorithm branch, multi-dimensional knowledge mining is performed on the first online business operation log debugging sample to obtain a third business log multi-dimensional knowledge vector;

[0022] The third business log multidimensional knowledge vector is optimized through the implicit algorithm branch to obtain the first business log multidimensional knowledge vector.

[0023] Optionally, optimizing the algorithm configuration variables of the multidimensional knowledge mining algorithm at least once based on the multivariate decision discrimination label includes:

[0024] When the debugging termination requirement is not met, the algorithm configuration variables of the implicit algorithm branches in the multidimensional knowledge mining algorithm are optimized according to the multivariate decision discrimination labels.

[0025] Optionally, the method further includes:

[0026] Acquire a fourth business log multi-dimensional knowledge vector of a second online business operation log debugging sample, wherein the second online business operation log debugging sample corresponds to the target business indicator keyword;

[0027] Determine a feature commonality weight between the fourth business log multidimensional knowledge vector and each attention embedded dense semantic vector in the attention feature relationship set to be debugged;

[0028] Determining a second target training error variable based on each of the feature commonality weights;

[0029] According to the second target training error variable, the attention feature relationship set to be debugged is optimized at least once to obtain the debugged attention feature relationship set.

[0030] Optionally, the second online business operation log debugging sample has a second associated attention dimension prior training annotation data; and determining the second target training error variable based on each of the feature commonality weights includes:

[0031] For each feature commonality weight, determining a third training error variable according to the feature commonality weight and the second associated attention dimension prior training annotation data;

[0032] Based on each of the third training error variables, the second target training error variable is determined.

[0033] Optionally, determining the second business log multidimensional knowledge vector according to the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector includes:

[0034] The first business log multidimensional knowledge vector and the target attention embedded dense semantic vector are integrated to obtain the second business log multidimensional knowledge vector.

[0035] The second aspect of an embodiment of the present application provides a warning service information generation system, comprising: a processor and a memory and a bus connected to the processor; the processor and the memory communicate with each other through the bus; the processor is used to call the computer program in the memory to execute the above-mentioned warning service information generation method based on multi-dimensional business monitoring.

[0036] According to a third aspect of an embodiment of the present application, a computer-readable storage medium is provided, on which a program is stored. When the program is executed by a processor, the method for generating early warning service information based on multi-dimensional business monitoring is implemented.

[0037] The method and system for generating early warning service information based on multi-dimensional business monitoring provided by the embodiment of the present application propose a new business monitoring method. By integrating multi-source data, the method provides a more comprehensive and multi-dimensional monitoring function, effectively making up for the shortcomings of traditional methods. At the same time, by using advanced multi-dimensional knowledge mining algorithms, the embodiment of the present application can deeply analyze and mine key information in log data to generate predictive early warning monitoring identification viewpoints. In this way, not only the real-time monitoring capability of the enterprise's business status is improved, but also the accuracy and efficiency of risk prevention and control are significantly enhanced, providing strong technical support for the stable operation of the enterprise. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. 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 related drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 A flowchart of a method for generating early warning service information based on multi-dimensional business monitoring provided in an embodiment of the present application.

[0040] Figure 2 A schematic diagram of a product module of a warning service information generation system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] The exemplary embodiments disclosed in the present application will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0042] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0043] See also Figure 1 , which is a flowchart of a method for generating early warning service information based on multi-dimensional business monitoring provided in an embodiment of the present application. The method is applied to an early warning service information generation system. The specific content description of the method includes steps 110 to 130.

[0044] Step 110: Obtain the online business operation log to be monitored for early warning.

[0045] Step 120 , by completing the debugged target multidimensional knowledge mining algorithm, multidimensional knowledge mining processing is performed on the online business operation log to be monitored for early warning, to obtain a target multidimensional knowledge vector.

[0046] Step 130: Obtain an early warning monitoring identification viewpoint of the online business operation log to be subjected to early warning monitoring based on the target multi-dimensional knowledge vector.

[0047] In an exemplary application scenario, the early warning service information generation system is deployed on a large e-commerce platform to monitor the business operation of the platform in real time and provide risk early warning services.

[0048] The early warning service information generation system will first collect the operation logs of online businesses in real time from various business modules of the e-commerce platform. These logs record various operations, events and status changes during the operation of the platform, and are an important data source for subsequent multi-dimensional knowledge mining and early warning monitoring.

[0049] After collecting the log data, the early warning service information generation system uses a pre-debugged multi-dimensional knowledge mining algorithm to perform in-depth processing on the log data. The algorithm can mine the logs from multiple dimensions (such as time, user behavior, transaction status, system performance, etc.), extract key information, and convert it into multi-dimensional knowledge vectors. These vectors not only contain the basic information in the logs, but also reflect the associations and patterns between the logs, providing a rich data foundation for subsequent early warning monitoring.

[0050] Based on the obtained multi-dimensional knowledge vectors, the early warning service information generation system further performs early warning monitoring and identification. The system analyzes and judges these vectors according to preset rules and models to identify possible risks and problems. For example, the system may detect an abnormal increase in transaction volume in a certain period of time, or an abnormal frequency of a certain user's behavior, etc. These are potential risk points. Once these risk points are identified, the system will generate corresponding early warning information. These early warning information not only includes the type and level of risk, but also provides possible causes and solutions. In this way, managers of e-commerce platforms can timely discover and deal with potential problems based on this information to ensure the stable operation of the platform.

[0051] It can be seen that the early warning service information generation system provides comprehensive and multi-dimensional business monitoring functions for e-commerce platforms by integrating multi-source data and using advanced multi-dimensional knowledge mining technology. This not only helps enterprises to discover and deal with potential problems in a timely manner, but also improves the operational efficiency and service quality of enterprises and reduces operational risks.

[0052] In combination with the above application scenarios, in step 110, the online business operation log to be monitored for early warning refers to the detailed log data that is generated in real time during the operation of the e-commerce platform and records the operation of the online business. These log data contain key contents such as operation information, transaction data, user behavior, system status changes, etc. of each business module of the platform during operation. These logs have not been analyzed and processed, but they are the original materials for the early warning service information generation system to carry out risk warning and monitoring. By deeply analyzing these logs, the system can identify abnormal situations in business operations, so as to provide timely early warning and response.

[0053] In the operation process of the early warning service information generation system, step 110 is the crucial first step. In this step, the core task of the system is to "obtain the online business operation log to be monitored for early warning". Specifically, the early warning service information generation system establishes data connection with each business module of the e-commerce platform, and collects and records the log data generated by these modules during operation in real time. These log data are real-time records of the business operation of the e-commerce platform, containing rich information, such as user login, browsing, purchase and other behaviors, commodity listing, price changes and other status, as well as system response time, error rate and other performance indicators. In order to ensure the integrity and real-time nature of the data, the early warning service information generation system adopts an efficient data collection mechanism, such as using a special log collection tool or interface to ensure that each log data can be captured and transmitted to the system in a timely and accurate manner. At the same time, the system will also perform preliminary processing and formatting on the collected log data to facilitate subsequent multi-dimensional knowledge mining and early warning monitoring analysis. This step provides a solid data foundation for subsequent risk early warning and monitoring. Only by obtaining comprehensive and accurate online business operation logs can the early warning service information generation system effectively conduct multi-dimensional knowledge mining, thereby identifying potential risks and problems and providing timely and effective risk early warning services for e-commerce platforms.

[0054] In step 120, the target multidimensional knowledge mining algorithm that has completed debugging refers to a multidimensional knowledge mining algorithm that has achieved the expected effect and is ready to be put into use in the early warning service information generation system after repeated testing and adjustment. This algorithm is a core component of the system, and its design purpose is to mine valuable multidimensional information from massive online business operation logs. Through this algorithm, the system can deeply analyze the log data and extract key information from multiple angles such as time, user behavior, transaction status, system performance, etc., thereby providing data support for risk warning and monitoring. The "complete debugging" status of the algorithm means that it has been rigorously tested and optimized to ensure accuracy and efficiency when processing actual log data.

[0055] In the operation process of the early warning service information generation system, step 120 is a key link, which involves in-depth processing and analysis of the collected online business operation logs. In this step, the system uses a multidimensional knowledge mining algorithm that has been debugged and optimized to process the log data to be monitored for early warning. The algorithm has a powerful data mining capability and can perform in-depth analysis of logs from multiple dimensions to extract key information. The multidimensional knowledge mining algorithm will identify and extract various important data features from the logs, such as user behavior patterns, transaction status changes, system performance indicators, etc. These data features not only reflect the real-time operation of the e-commerce platform business, but may also hide potential risks and problems. Through the processing of the algorithm, these key information will be converted into a special data structure: the target multidimensional knowledge vector. These vectors are abstract representations of log data. They not only contain the basic information in the log, but more importantly, reveal the internal connections and patterns between the log data. These vectors will become the basis for subsequent early warning monitoring and analysis. Step 120 provides a rich data foundation for the early warning service information generation system, enabling the system to more accurately identify abnormal situations in business operations and issue early warnings in a timely manner. At the same time, this step also provides e-commerce platform managers with deeper business insights, helping them better understand platform operations and develop effective risk management strategies.

[0056] It is worth mentioning that in the above technical solution, the target multidimensional knowledge vector is the core information extracted from the online business operation log to be monitored for early warning through multidimensional knowledge mining. These vectors are represented in the form of numerical features and can fully reflect the multidimensional information and potential patterns in the log data. The following is a detailed introduction through a specific numerical feature vector example.

[0057] (1) Composition of numerical feature vectors:

[0058] Each numerical feature vector can consist of multiple elements, each of which represents a specific data feature or indicator.

[0059] For example, a numerical feature vector may contain data on multiple dimensions, such as user login frequency, transaction amount, page visit duration, system response time, etc.

[0060] Example: User behavior analysis numerical feature vector

[0061] There is a numerical feature vector for user behavior, which may contain the following elements:

[0062] User daily activity: indicates the number of times a user is active every day, such as [0.8, 1.2, 1.0, ...] (the values ​​are examples and have no units).

[0063] Average visit duration: The average duration of each visit by a user, in minutes, such as [5.3, 6.2, 5.8, ...].

[0064] Purchase conversion rate: the proportion of users who actually purchase products after browsing them, such as [0.03, 0.05, 0.04, ...].

[0065] (3) The meaning and application of vectors:

[0066] These numerical feature vectors not only reflect the individual behavior of users, but also reveal the behavioral patterns and trends of user groups.

[0067] By analyzing these vectors, abnormal user behaviors, such as a sudden increase in activity or an abnormally low purchase conversion rate, can be detected in a timely manner, thereby triggering an early warning mechanism.

[0068] (4) Dynamicity and timeliness:

[0069] The numerical feature vector is dynamic and is updated as user behavior changes.

[0070] This enables the system to monitor business status in real time and respond quickly to emergencies.

[0071] (5) Integration of multi-dimensional information:

[0072] In addition to user behavior, numerical feature vectors can also contain information on multiple dimensions such as transaction status and system performance.

[0073] The integration of this information enables the system to comprehensively evaluate business operations from multiple perspectives and improve the accuracy and timeliness of early warnings.

[0074] In summary, the target multi-dimensional knowledge vector, as the core invention of this application, comprehensively and dynamically reflects the multi-dimensional information in the online business operation log in the form of numerical features. These vectors play a key role in the early warning monitoring system and provide strong data support for timely discovery of potential risks and problems.

[0075] In step 130, the early warning monitoring identification viewpoint refers to the conclusion or judgment drawn by the early warning service information generation system based on the multi-dimensional knowledge vector analysis. These viewpoints are the identification and interpretation of potential risks or abnormal situations in the online business operation logs, and they provide managers with key information about the business operation status of the e-commerce platform. Early warning monitoring identification viewpoints may include but are not limited to the identification and early warning of abnormal transaction volume, changes in user behavior patterns, system performance degradation, etc., and are important decision-making support to ensure the stable operation of the e-commerce platform.

[0076] In the operation process of the early warning service information generation system, step 130 is the key link in forming the final early warning information. In this step, the system conducts in-depth analysis and judgment based on the target multidimensional knowledge vector obtained in the previous step to form an early warning monitoring and identification viewpoint. First, the system checks each element in the multidimensional knowledge vector one by one according to the preset rules and models. These rules and models are built based on historical data and expert knowledge, and they can help the system identify abnormal patterns and potential risks in log data. Then, the system conducts a comprehensive evaluation of these abnormal patterns and potential risks to determine their possible impact on the business operation of the e-commerce platform. In this process, the system considers multiple factors, such as the duration of the abnormality, the scope of impact, and the possible consequences. Finally, based on the above analysis and evaluation, the system forms specific early warning monitoring and identification views. These views may include: abnormal increase in transaction volume, there may be short-term load pressure caused by brushing orders or promotional activities; abnormal frequency of a certain user behavior may indicate malicious behavior or system vulnerabilities being exploited; system performance is degraded, and hardware upgrades or optimized software configurations may be required. These early warning monitoring and identification viewpoints will be presented to the managers of the e-commerce platform in a clear and explicit manner, helping them to quickly understand the risks and problems in the current business operation and take corresponding countermeasures. In this way, the early warning service information generation system not only improves the risk response ability of the e-commerce platform, but also provides a strong guarantee for the stable operation of the enterprise.

[0077] It is worth mentioning that the early warning monitoring identification viewpoint is the core output of the early warning service information generation system after analyzing the online business operation log. It represents the system's judgment and identification of potential risks or abnormal situations. These views are derived from the target multidimensional knowledge vector after multidimensional knowledge mining processing, and they provide key decision support and risk indication for managers of e-commerce platforms.

[0078] (1) Formation process

[0079] The formation of early warning monitoring and identification viewpoints is a complex process involving the comprehensive application of multiple steps and technologies. First, the system processes massive online business operation logs through multi-dimensional knowledge mining algorithms, extracts key information and converts it into multi-dimensional knowledge vectors. These vectors contain multiple dimensional information in the log data, such as time, user behavior, transaction status, system performance, etc.

[0080] Next, the system uses preset rules and models to conduct in-depth analysis of these multi-dimensional knowledge vectors. These rules and models are built based on historical data, industry standards, and expert knowledge to help the system accurately identify abnormal patterns and potential risks. For example, the system may detect a sudden surge in transaction volume or an unusually frequent user login behavior, which are signals that require further attention.

[0081] (2) Content interpretation

[0082] The content of early warning monitoring and identification viewpoints is rich and varied, depending on the multi-dimensional knowledge vector analyzed and the preset rule model. The following are some possible examples of early warning monitoring and identification viewpoints and their interpretations:

[0083] Abnormally high trading volume: This type of view is generated when the system detects that the trading volume in a certain period of time is far higher than the historical average. This may mean that there are promotions or special events that lead to active trading, but it may also hide risks such as fake orders and fraud.

[0084] Abnormal user behavior: If the frequency and pattern of user login, browsing or purchasing behavior are inconsistent with the norm, the system will issue such an alert. This may indicate malicious behavior, account compromise or system vulnerability exploitation.

[0085] System performance degradation: When system response time increases and error rates rise, it indicates that the platform may face technical challenges. Such observations prompt managers to check whether hardware resources, network bandwidth or software configuration need to be optimized or upgraded.

[0086] Unusual transaction patterns: For example, a large number of small transactions or frequent refunds may indicate fraudulent activity or product quality issues.

[0087] Anomalies in region or device: If a large number of transactions suddenly come from an atypical region or device type, this may indicate a security risk.

[0088] (3) Application value

[0089] The application value of early warning monitoring and identification is mainly reflected in the following aspects:

[0090] Risk prevention: By identifying potential risks in a timely manner, e-commerce platforms can take preventive measures before problems occur, thereby avoiding or reducing losses.

[0091] Improved operational efficiency: Understanding anomalies in system performance and user behavior helps the platform optimize operational strategies and improve service quality and user experience.

[0092] Security assurance: Rapid identification of abnormal logins, transactions and other behaviors helps enhance the security of the platform and protect user data and assets.

[0093] In summary, the early warning monitoring and identification viewpoint is one of the core outputs of the early warning service information generation system. It provides comprehensive risk warning and decision-making support for e-commerce platforms and is a key link in ensuring the stable and safe operation of the platform.

[0094] The embodiment of the present application realizes a comprehensive and multi-dimensional business monitoring function by integrating multi-source data, significantly improving the enterprise's real-time monitoring ability of the business operation status. The system can not only efficiently collect and analyze the operation logs of online businesses, but also use advanced multi-dimensional knowledge mining algorithms to extract key target multi-dimensional knowledge vectors from these logs. These vectors not only fully reflect the operation status of the business, but also accurately reveal potential risks and problems.

[0095] Furthermore, by intelligently analyzing these target multi-dimensional knowledge vectors, the system can generate early warning monitoring and identification viewpoints that are highly targeted and predictive, thereby helping enterprises to promptly discover and solve potential problems. This innovative function greatly enhances the enterprise's ability to prevent and control operational risks and ensures the stability and continuity of business operations. At the same time, the system also supports customized log data filtering, which meets the personalized analysis needs of enterprises and further improves operational efficiency and risk management levels. In general, the embodiments of the present application provide enterprises with an efficient and intelligent business monitoring solution that helps reduce operational risks and ensure the continued healthy development of the business.

[0096] In some optional embodiments, the method further includes: using a multidimensional knowledge mining algorithm to be debugged, performing multidimensional knowledge mining processing on a first online business operation log debugging sample to obtain a first business log multidimensional knowledge vector, wherein the first online business operation log debugging sample corresponds to a target business indicator keyword; determining a second business log multidimensional knowledge vector based on the first business log multidimensional knowledge vector and a target attention embedding dense semantic vector, wherein the target attention embedding dense semantic vector is determined from an attention feature relationship set that has completed debugging, wherein the attention feature relationship set that has completed debugging includes an attention embedding dense semantic vector corresponding to at least one associated attention dimension, each of which respectively represents a monitoring and early warning description vector of the online business operation log corresponding to the target business indicator keyword in the corresponding associated attention dimension; using a multivariate decision tree algorithm, performing multivariate decision discrimination on the second business log multidimensional knowledge vector to obtain a multivariate decision discrimination label; optimizing the algorithm configuration variables of the multidimensional knowledge mining algorithm at least once based on the multivariate decision discrimination label to obtain the debugged multidimensional knowledge mining algorithm.

[0097] In this embodiment, the early warning service information generation system also includes a key algorithm debugging and optimization link in the process of performing its core business: providing real-time monitoring and early warning services for e-commerce platforms. This link is designed to ensure the accuracy and efficiency of the multi-dimensional knowledge mining algorithm, so as to more accurately identify potential risks in online business operation logs.

[0098] First, the system selects a set of specific first online business operation log debugging samples that are closely related to the target business indicator keywords. These keywords may involve core indicators of e-commerce platforms such as transaction volume, user activity, and system response time. Using the multidimensional knowledge mining algorithm to be debugged, the system deeply processes these debugging samples, extracts key information from them, and forms multidimensional knowledge vectors of the first business log. These vectors are abstract representations of log data, and they capture the multidimensional features and associations in the logs.

[0099] Next, the system determines the target attention embedding dense semantic vector based on the debugged attention feature relationship set. The relationship set contains the attention embedding dense semantic vectors corresponding to multiple associated attention dimensions. Each vector represents the monitoring and warning description of the online business operation log related to the target business indicator keyword in the corresponding associated attention dimension. The system selects the attention embedding dense semantic vector that is most relevant to the current business indicator to ensure the accuracy and pertinence of the monitoring and warning.

[0100] With the target attention embedded dense semantic vector, the system can further determine the second business log multidimensional knowledge vector. This vector combines the information of the first business log multidimensional knowledge vector and the content of the target attention embedded dense semantic vector, providing a richer data foundation for the subsequent multivariate decision tree algorithm.

[0101] Subsequently, the warning service information generation system uses a multivariate decision tree algorithm to perform multivariate decision discrimination on the multidimensional knowledge vector of the second business log. The algorithm can classify and discriminate according to the multidimensional characteristics of the vector, and assign a multivariate decision discrimination label to each vector. These labels reflect the potential risks and warning levels contained in the log data.

[0102] Finally, the system optimizes the algorithm configuration variables of the multidimensional knowledge mining algorithm based on these multivariate decision discrimination labels. This optimization process may involve adjusting the algorithm parameters, improving the feature extraction method, or optimizing the model structure. Through at least one optimization iteration, the system finally obtains the debugged multidimensional knowledge mining algorithm, which will be more accurate and efficient in subsequent real-time monitoring and early warning services.

[0103] Designed in this way, the early warning service information generation system not only improves the accuracy and efficiency of the multidimensional knowledge mining algorithm, but also enhances its ability to identify potential risks in online business operation logs. This optimized algorithm can more accurately capture the multidimensional features and associations in the log data, thereby generating more reliable early warning monitoring and identification viewpoints. For e-commerce platforms, this means that potential problems can be discovered and responded to more quickly, ensuring the stable operation of the platform, reducing operational risks, and improving user experience. Therefore, this embodiment not only improves the intelligence level of the early warning service, but also provides strong technical support for the risk management of enterprises.

[0104] It can be understood that, combined with the above content, the debugged multidimensional knowledge mining algorithm plays a core role in the early warning service information generation system. The algorithm integrates and optimizes multiple neural network algorithms to achieve in-depth analysis and key information extraction of online business operation logs. The following will introduce three specific neural network algorithms in detail, which play an important role in the debugged multidimensional knowledge mining algorithm.

[0105] (1) Convolutional Neural Network (CNN):

[0106] CNN is a deep learning model that is particularly suitable for processing images and grid structured data, but in multidimensional knowledge mining, it can also be used to extract local features in log data.

[0107] During the debugging process, the convolutional and pooling layers of CNN were optimized to make them more adaptable to the characteristics of log data.

[0108] Through convolution operations, CNN can capture local patterns in log data and reduce the data dimension through pooling operations to extract key features.

[0109] In the multidimensional knowledge mining algorithm, CNN helps to effectively extract valuable information from large amounts of log data, providing an important basis for subsequent early warning analysis.

[0110] (2) Recurrent Neural Network (RNN):

[0111] RNN is suitable for processing sequential data, such as time series and text, and it is able to consider the contextual information of the data.

[0112] During the debugging process, the RNN was improved based on the temporal characteristics of log data so that it can better capture the long-term dependencies in the log sequence.

[0113] RNN is used in multidimensional knowledge mining algorithms to analyze the time series characteristics of log data and help the system identify abnormal time patterns or periodic behaviors.

[0114] Through RNN processing, potential risk points can be predicted and identified more accurately, improving the accuracy and timeliness of early warning.

[0115] (3) Autoencoder:

[0116] An autoencoder is an unsupervised learning model used to learn efficient representations of data.

[0117] During the debugging process, the encoder and decoder of the autoencoder were optimized to better adapt to the characteristics of log data.

[0118] Autoencoders are used for data dimensionality reduction and feature extraction in multidimensional knowledge mining algorithms. They can extract key features from complex log data.

[0119] Through the processing of autoencoders, log data can be compressed and represented more effectively, thereby improving the efficiency and accuracy of multidimensional knowledge mining.

[0120] In summary, the debugged multi-dimensional knowledge mining algorithm integrates and optimizes neural network algorithms such as convolutional neural networks, recurrent neural networks, and autoencoders to achieve in-depth analysis of online business operation logs and extraction of key information. The application of this algorithm not only improves the intelligence level of early warning services, but also provides strong technical support for corporate risk management.

[0121] In other preferred embodiments, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once based on the multivariate decision discrimination label to obtain the debugged multidimensional knowledge mining algorithm, including: determining a first target training error variable based on the multivariate decision discrimination label and set prior training annotation data; based on the first target training error variable, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once to obtain the debugged multidimensional knowledge mining algorithm.

[0122] Based on this embodiment, the early warning service information generation system performs optimization work on the multi-dimensional knowledge mining algorithm. The core of this process is to use multi-decision discrimination labels.

[0123] First, the warning service information generation system obtains multiple decision-making discrimination labels, which are based on a large amount of data analysis and expert experience, and can accurately reflect the decision-making tendencies in different situations. At the same time, the system will also set or obtain a set of prior training annotation data, which are verified and representative sample data for training and optimizing algorithms.

[0124] Next, the system determines an indicator called the first target training error variable based on the multivariate decision discriminant label and the prior training annotated data. This error variable is an important parameter to measure the difference between the algorithm's prediction results and the actual annotated data. By minimizing this error variable, the algorithm can be guided to learn and predict more accurately.

[0125] After determining the first target training error variable, the early warning service information generation system optimizes the algorithm configuration variables of the multidimensional knowledge mining algorithm. The algorithm configuration variables here may include learning rate, regularization parameters, number of iterations, etc., which directly affect the performance and accuracy of the algorithm. The optimization process may involve multiple iterations and adjustments to ensure that the algorithm can achieve the best performance in various situations.

[0126] Through at least one such optimization process, the early warning service information generation system will eventually obtain a debugged multi-dimensional knowledge mining algorithm. When processing complex data, the optimized algorithm will be able to extract useful information more accurately and provide more reliable data support for subsequent early warning services.

[0127] In summary, by optimizing the multidimensional knowledge mining algorithm based on multi-decision discrimination labels, the early warning service information generation system can significantly improve the accuracy and efficiency of the algorithm. This not only helps the system to more accurately identify and predict potential risk points, but also provides a more scientific and reliable basis for relevant decisions. This optimization method is of great significance in improving the quality and effectiveness of early warning services, and lays a solid foundation for improving the performance of early warning service information generation systems in practical applications.

[0128] In some examples, the set prior training annotation data includes the prior keyword annotation data of the target business indicator keyword and the first associated attention dimension prior training annotation data of the target attention embedded dense semantic vector, and the multi-decision discrimination label includes business indicator keyword data and log associated attention dimension information; determining the first target training error variable based on the multi-decision discrimination label and the set prior training annotation data includes: determining the first training error variable based on the business indicator keyword data and the prior keyword annotation data; determining the second training error variable based on the log associated attention dimension information and the first associated attention dimension prior training annotation data; determining the first target training error variable based on the first training error variable and the second training error variable.

[0129] In this example, the warning service information generation system performs a series of complex operations to determine the first target training error variable, which involves the comprehensive application of multivariate decision discriminant labels and set prior training annotation data. The following is a detailed example explanation of this process.

[0130] First, the prior training annotation data set by the system includes two parts: one is the prior keyword annotation data of the target business indicator keywords. These data provide the system with accurate annotations of business indicator keywords and are an important reference in the training process. The other is the first-association attention dimension prior training annotation data of the target attention embedded dense semantic vector. These data reflect the annotation information on a specific attention dimension, which helps the system better understand and process complex semantic relationships.

[0131] At the same time, the multivariate decision discriminant label also includes two parts: business indicator keyword data and log-related attention dimension information. Business indicator keyword data provides the system with actual business indicator keyword samples, while log-related attention dimension information reflects which attention dimensions are closely related to business indicators in the actual log data.

[0132] In the process of determining the first target training error variable, the early warning service information generation system first calculates the first training error variable based on the business indicator keyword data and the prior keyword annotation data. This error variable reflects the difference between the business indicator keyword predicted by the system and the actual annotation data, and is an important indicator for evaluating the accuracy of the system prediction.

[0133] Next, the system determines the second training error variable based on the log-related attention dimension information and the first-related attention dimension prior training annotation data. This error variable measures the accuracy of the system in processing the log-related attention dimension, which helps the system better capture and understand the key information in the log data.

[0134] Finally, the warning service information generation system combines the first training error variable and the second training error variable to determine the first target training error variable. This comprehensive error variable fully reflects the overall performance of the system in processing business indicator keywords and log-related attention dimensions, and is an important basis for subsequent optimization algorithm configuration variables.

[0135] Through this process, the early warning service information generation system can more comprehensively evaluate its own prediction performance and effectively optimize existing problems. This not only improves the prediction accuracy of the system, but also makes the system more flexible and reliable in practical applications, and can better meet the needs of users.

[0136] In summary, by comprehensively applying the multivariate decision discriminant labels and the set prior training annotation data to determine the first target training error variable, the early warning service information generation system can more accurately evaluate its own performance and optimize it in a targeted manner. This technical solution significantly improves the prediction accuracy and reliability of the system, providing a strong guarantee for the accuracy and practicality of the early warning service.

[0137] In some examples, the multidimensional knowledge mining algorithm includes a pyramid algorithm branch and an implicit algorithm branch; the multidimensional knowledge mining process is performed on the first online business operation log debugging sample using the multidimensional knowledge mining algorithm to be debugged to obtain a first business log multidimensional knowledge vector, including: performing multidimensional knowledge mining process on the first online business operation log debugging sample through the pyramid algorithm branch to obtain a third business log multidimensional knowledge vector; optimizing the third business log multidimensional knowledge vector through the implicit algorithm branch to obtain the first business log multidimensional knowledge vector.

[0138] Based on this example, the early warning service information generation system uses a special multi-dimensional knowledge mining algorithm, which contains two important branches: the pyramid algorithm branch and the implicit algorithm branch. These two branches play different roles in the algorithm and work together to complete the in-depth analysis and knowledge extraction of online business operation logs.

[0139] First, the early warning service information generation system uses the pyramid algorithm branch to perform multi-dimensional knowledge mining on the first online business operation log debugging sample. The pyramid algorithm branch is known for its hierarchical processing method, which can gradually and deeply analyze the log data to extract key information and features. In this process, the pyramid algorithm branch will identify and organize various elements in the log, such as timestamps, user behaviors, system events, etc., and convert this information into a preliminary multi-dimensional knowledge vector, which is called the third business log multi-dimensional knowledge vector.

[0140] Next, the implicit algorithm branch will optimize the third business log multidimensional knowledge vector. The implicit algorithm branch is designed to further refine and enhance the output of the pyramid algorithm branch. It uses complex mathematical models and calculation methods to fine-tune the third business log multidimensional knowledge vector to remove redundant information, strengthen key features, and reveal implicit patterns and associations in the log data. After being processed by the implicit algorithm branch, the third business log multidimensional knowledge vector is optimized into a more accurate and comprehensive first business log multidimensional knowledge vector.

[0141] In this way, the early warning service information generation system can have a deeper understanding of the internal structure and dynamic changes of the online business operation log. Through the synergy of the pyramid algorithm branch and the implicit algorithm branch, the system can capture the subtle differences and abnormal patterns in the log data, thereby providing more accurate and timely information support for subsequent early warning services. In detail, by using a multidimensional knowledge mining algorithm containing a pyramid algorithm branch and an implicit algorithm branch, the early warning service information generation system can achieve efficient analysis and knowledge extraction of online business operation logs. This technical solution not only improves the system's processing capabilities and depth of understanding of log data, but also provides a strong guarantee for the accuracy and real-time nature of the early warning service. By continuously optimizing and improving the multidimensional knowledge mining algorithm, the early warning service information generation system will be able to better serve various online businesses and ensure the stable operation and continuous improvement of the business.

[0142] In a replaceable embodiment, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once based on the multivariate decision discrimination label, including: when the debugging termination requirements are not met, the algorithm configuration variables of the implicit algorithm branches in the multidimensional knowledge mining algorithm are optimized based on the multivariate decision discrimination label.

[0143] In this embodiment, the warning service information generation system adopts a specific optimization strategy, that is, based on the multi-decision discrimination label, the algorithm configuration variables of the implicit algorithm branch in the multi-dimensional knowledge mining algorithm are optimized. This optimization process is performed without meeting the debugging termination requirements, aiming to further improve the performance and accuracy of the algorithm.

[0144] Specifically, when the system finds that the performance of the multidimensional knowledge mining algorithm does not meet expectations or there is room for further optimization, this optimization process will be triggered. The system will first analyze the multi-decision discrimination labels, which contain rich decision information and discrimination basis, and are an important reference in the optimization process.

[0145] Next, the system focuses on the implicit algorithm branch in the multidimensional knowledge mining algorithm. The implicit algorithm branch plays a key role in the algorithm, and its performance directly affects the effect of the entire algorithm. Therefore, the system carefully adjusts and optimizes the algorithm configuration variables of this branch.

[0146] During the optimization process, the system tries different combinations of configuration variables to find the best performance. These configuration variables may include key parameters such as learning rate, regularization coefficient, and number of iterations. The system continuously adjusts these parameters based on the feedback from the multivariate decision discriminant labels in order to achieve better mining results.

[0147] Through this optimization process, the early warning service information generation system can specifically improve the performance of implicit algorithm branches in the multidimensional knowledge mining algorithm, thereby improving the accuracy and efficiency of the algorithm as a whole. This optimization strategy not only helps the system better adapt to the complex and changing data environment, but also provides more accurate and reliable information support for subsequent early warning services.

[0148] In this way, by optimizing the implicit algorithm branches in the multidimensional knowledge mining algorithm based on the multi-decision discrimination labels when the debugging termination requirements are not met, the early warning service information generation system can achieve continuous improvement in algorithm performance. This technical solution provides the system with a powerful self-optimization capability, enabling it to continuously evolve and improve in practical applications, and better serve the generation and delivery of early warning information.

[0149] In some further optional embodiments, the method further includes: obtaining a fourth business log multidimensional knowledge vector of a second online business operation log debugging sample, wherein the second online business operation log debugging sample corresponds to the target business indicator keyword; determining a feature commonality weight between the fourth business log multidimensional knowledge vector and each attention embedded dense semantic vector in the attention feature relationship set to be debugged; determining a second target training error variable based on each of the feature commonality weights; and optimizing the attention feature relationship set to be debugged at least once based on the second target training error variable to obtain the debugged attention feature relationship set.

[0150] Based on this embodiment, the warning service information generation system performs a series of complex operations to optimize the attention feature relationship set, thereby improving the accuracy and effectiveness of the warning information. The following is a detailed example explanation of this process.

[0151] First, the system obtains the fourth business log multidimensional knowledge vector of the second online business operation log debugging sample. This vector is extracted from the second online business operation log debugging sample by a multidimensional knowledge mining algorithm, and it contains key information and features in the log. At the same time, the debugging sample corresponds to the target business indicator keyword, which means that it is closely related to the specific business indicator that the system focuses on.

[0152] Next, the system determines the feature commonality weights between the fourth business log multidimensional knowledge vector and each attention embedded dense semantic vector in the attention feature relationship set to be debugged. The purpose of this step is to find the common features between the log multidimensional knowledge vector and the attention embedded dense semantic vector and quantify the similarity between them. The determination of feature commonality weights helps the system understand the relationship between log data and attention features more accurately.

[0153] Then, based on each feature commonality weight, the system determines the second target training error variable. This error variable reflects the difference between the attention feature predicted by the system and the actual log data. By minimizing this error variable, the system can optimize the attention feature relationship set to make it closer to the actual business log data.

[0154] Finally, based on the second target training error variable, the system optimizes the attention feature relationship set to be debugged at least once. The optimization process may include adjusting the parameters of the attention embedding dense semantic vector, changing the connection mode of the feature relationship, etc. Through continuous optimization, the system can obtain a debugged attention feature relationship set, which can more accurately reflect the key information and features in the business log data.

[0155] In this way, the early warning service information generation system can more accurately capture and understand the key information in the business log data, thereby improving the accuracy and effectiveness of the early warning information. At the same time, through the optimization of the attention feature relationship set, the system can also better adapt to different business scenarios and needs, and improve the flexibility and practicality of the early warning service. In other words, by obtaining the multi-dimensional knowledge vector of the fourth business log of the second online business operation log debugging sample, and optimizing the attention feature relationship set according to the feature commonality weight, the early warning service information generation system can significantly improve the accuracy and effectiveness of the early warning information. This technical solution provides strong support for the continuous improvement and enhancement of the early warning service, and helps to ensure the stable operation of the business and the security of customer assets.

[0156] In some possible embodiments, the second online business operation log debugging sample has a second associated attention dimension prior training annotation data; determining the second target training error variable based on each of the feature commonality weights includes: for each feature commonality weight, determining a third training error variable according to the feature commonality weight and the second associated attention dimension prior training annotation data; determining the second target training error variable based on each of the third training error variables.

[0157] In this embodiment, the warning service information generation system processes the second online business operation log debugging samples with special attributes. These samples not only contain the key data of business operation, but also have the second associated attention dimension prior training annotation data associated with them. These data provide additional training guidance for the system, which helps to improve the accuracy and effectiveness of the warning information.

[0158] When processing these debugging samples, the system first obtains the commonality weight of each feature, which is obtained by comparing the multi-dimensional knowledge vector of the fourth business log with the attention embedding dense semantic vector in the attention feature relationship set. These feature commonality weights reflect the similarity and correlation between the log data and the attention feature.

[0159] Next, the system uses these feature commonality weights and the prior training annotation data of the second associated attention dimension to determine the third training error variable. This error variable actually measures the difference between the attention features predicted by the system and the actual annotated data. By calculating this difference, the system can clearly know in which aspects the prediction is biased, thus providing direction for subsequent optimization.

[0160] After determining the third training error variables, the system further determines the second target training error variable based on these variables. This variable integrates the information of all third training error variables and provides a global training error indicator for the system. Through this indicator, the system can evaluate the accuracy of the current attention feature relationship set and optimize accordingly.

[0161] The optimization process may include adjusting the parameters of the attention embedding dense semantic vector, changing the connection mode of feature relations, or introducing new features, etc. The system continuously iterates and optimizes according to the feedback of the second target training error variable until satisfactory performance is achieved.

[0162] In this way, the warning service information generation system can make full use of the prior training annotation data to more accurately capture and understand the key information in the business log data. At the same time, by continuously iterating and optimizing the attention feature relationship set, the system can improve the accuracy and effectiveness of the warning information and better adapt to different business scenarios and needs. In other words, by combining the feature commonality weights and the second related attention dimension prior training annotation data to determine the second target training error variable, the warning service information generation system can significantly improve the accuracy and effectiveness of the warning information. This technical solution not only enhances the system's self-learning and optimization capabilities, but also provides strong support for the continuous improvement and enhancement of warning services, thereby better ensuring the stable operation of the business and the safety of customer assets.

[0163] In other optional embodiments, determining the second business log multidimensional knowledge vector based on the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector includes: integrating the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector to obtain the second business log multidimensional knowledge vector.

[0164] In this embodiment, the warning service information generation system adopts a specific method to determine the second service log multidimensional knowledge vector. The process involves integrating the first service log multidimensional knowledge vector with the target attention embedding dense semantic vector.

[0165] Specifically, the system first obtains the multidimensional knowledge vector of the first business log, which is extracted from the first online business operation log debugging sample through a multidimensional knowledge mining algorithm and contains key information and features in the log. At the same time, the system also obtains the target attention embedded dense semantic vector, which represents the specific attention features that the system needs to pay attention to.

[0166] Next, the system integrates the first business log multidimensional knowledge vector with the target attention embedded dense semantic vector. The integration process may be achieved through some mathematical operation or model, such as weighted summation, splicing, or fusion through a neural network. The purpose of the integration is to embed the target attention features into the first business log multidimensional knowledge vector, thereby obtaining a new vector that incorporates specific attention information. The new vector can be understood as the second business log multidimensional knowledge vector. It not only contains the key information and features in the original log data, but also incorporates the specific attention features that the system pays attention to. This enables the second business log multidimensional knowledge vector to more comprehensively reflect the inherent structure and dynamic changes of the log data, while highlighting the key information that the system cares about.

[0167] Through the implementation of this technical solution, the warning service information generation system can more accurately capture and understand the key information in the business log data, especially those related to the attention features that the system focuses on. This helps to improve the accuracy and effectiveness of warning information, enabling the system to respond to potential problems and risks more quickly. In this way, by integrating the first business log multidimensional knowledge vector with the target attention embedded dense semantic vector, the warning service information generation system can generate a more comprehensive and accurate second business log multidimensional knowledge vector. This technical solution enhances the system's processing capabilities and depth of understanding of log data, providing a strong guarantee for the accuracy and real-time nature of the warning service.

[0168] Please refer to Figure 2 The embodiment of the present application also provides a warning service information generation system 100, including a processor 111, and a memory 112 and a bus 113 connected to the processor 111. The processor 111 and the memory 112 communicate with each other through the bus 113. The processor 111 is used to call the program instructions in the memory 112 to execute the above-mentioned warning service information generation method based on multi-dimensional business monitoring.

[0169] In summary, by integrating multi-source data, comprehensive and multi-dimensional business monitoring functions are realized, significantly improving the company's real-time monitoring capabilities for business operation status. The system can not only efficiently collect and analyze online business operation logs, but also use advanced multi-dimensional knowledge mining algorithms to extract key target multi-dimensional knowledge vectors from these logs. These vectors not only fully reflect the operation status of the business, but also accurately reveal potential risks and problems.

[0170] Furthermore, by intelligently analyzing these target multi-dimensional knowledge vectors, the system can generate early warning monitoring and identification viewpoints that are highly targeted and predictive, thereby helping enterprises to promptly discover and solve potential problems. This innovative function greatly enhances the enterprise's ability to prevent and control operational risks and ensures the stability and continuity of business operations. At the same time, the system also supports customized log data filtering, which meets the personalized analysis needs of enterprises and further improves operational efficiency and risk management levels. In general, the embodiments of the present application provide enterprises with an efficient and intelligent business monitoring solution that helps reduce operational risks and ensure the continued healthy development of the business.

[0171] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity or system including the element.

[0172] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0173] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for generating early warning service information based on multi-dimensional business monitoring, characterized in that: Applied to a warning service information generation system, the method comprises: Obtain online business operation logs to be monitored for early warning; By completing the debugging of the target multidimensional knowledge mining algorithm, multidimensional knowledge mining processing is performed on the online business operation log to be monitored for early warning, so as to obtain a target multidimensional knowledge vector; According to the target multi-dimensional knowledge vector, an early warning monitoring identification viewpoint of the online business operation log to be subjected to early warning monitoring is obtained.

2. The method according to claim 1, characterized in that The method further comprises: Using the multidimensional knowledge mining algorithm to be debugged, multidimensional knowledge mining processing is performed on the first online business operation log debugging sample to obtain a first business log multidimensional knowledge vector, wherein the first online business operation log debugging sample corresponds to a target business indicator keyword; Determine a second business log multidimensional knowledge vector according to the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector, wherein the target attention embedded dense semantic vector is determined from the attention feature relationship set of the completed debugging, wherein the attention feature relationship set of the completed debugging includes at least one attention embedded dense semantic vector corresponding to each associated attention dimension, and each of the attention embedded dense semantic vectors respectively represents a monitoring warning description vector of the online business operation log corresponding to the target business indicator keyword in the corresponding associated attention dimension; Using a multivariate decision tree algorithm, performing multivariate decision discrimination on the multidimensional knowledge vector of the second business log to obtain a multivariate decision discrimination label; According to the multivariate decision discrimination label, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once to obtain the debugged multidimensional knowledge mining algorithm.

3. The method according to claim 2, characterized in that The method of optimizing the algorithm configuration variables of the multidimensional knowledge mining algorithm at least once based on the multivariate decision discrimination label to obtain the debugged multidimensional knowledge mining algorithm includes: Determining a first target training error variable according to the multivariate decision discriminant label and the set prior training annotation data; According to the first target training error variable, the algorithm configuration variables of the multidimensional knowledge mining algorithm are optimized at least once to obtain the debugged multidimensional knowledge mining algorithm.

4. The method according to claim 3, characterized in that The set prior training annotation data includes the prior keyword annotation data of the target business indicator keyword and the first associated attention dimension prior training annotation data of the target attention embedded dense semantic vector, and the multivariate decision discrimination label includes the business indicator keyword data and the log associated attention dimension information; The step of determining a first target training error variable based on the multivariate decision discriminant label and the set prior training annotation data includes: Determining a first training error variable according to the business indicator keyword data and the prior keyword labeling data; Determining a second training error variable according to the log-associated attention dimension information and the first-associated attention dimension prior training annotation data; The first target training error variable is determined according to the first training error variable and the second training error variable.

5. The method according to claim 2, characterized in that: The multidimensional knowledge mining algorithm includes a pyramid algorithm branch and an implicit algorithm branch; the multidimensional knowledge mining algorithm to be debugged is used to perform multidimensional knowledge mining processing on the first online business operation log debugging sample to obtain a multidimensional knowledge vector of the first business log, including: Through the pyramid algorithm branch, multi-dimensional knowledge mining is performed on the first online business operation log debugging sample to obtain a third business log multi-dimensional knowledge vector; The third business log multidimensional knowledge vector is optimized through the implicit algorithm branch to obtain the first business log multidimensional knowledge vector.

6. The method according to claim 5, characterized in that The step of optimizing the algorithm configuration variables of the multidimensional knowledge mining algorithm at least once based on the multivariate decision discrimination labels includes: When the debugging termination requirement is not met, the algorithm configuration variables of the implicit algorithm branches in the multidimensional knowledge mining algorithm are optimized according to the multivariate decision discrimination labels.

7. The method according to claim 2, characterized in that The method further comprises: Acquire a fourth business log multi-dimensional knowledge vector of a second online business operation log debugging sample, wherein the second online business operation log debugging sample corresponds to the target business indicator keyword; Determine a feature commonality weight between the fourth business log multidimensional knowledge vector and each attention embedded dense semantic vector in the attention feature relationship set to be debugged; Determining a second target training error variable based on each of the feature commonality weights; According to the second target training error variable, optimizing the attention feature relationship set to be debugged at least once to obtain the debugged attention feature relationship set; The second online business operation log debugging sample has a second associated attention dimension prior training annotation data; and determining the second target training error variable based on each of the feature commonality weights includes: For each feature commonality weight, determining a third training error variable according to the feature commonality weight and the second associated attention dimension prior training annotation data; Based on each of the third training error variables, the second target training error variable is determined.

8. The method according to any one of claims 2 to 7, characterized in that The determining of the second business log multidimensional knowledge vector according to the first business log multidimensional knowledge vector and the target attention embedded dense semantic vector includes: The first business log multidimensional knowledge vector and the target attention embedded dense semantic vector are integrated to obtain the second business log multidimensional knowledge vector.

9. A warning service information generation system, characterized in that: It comprises a processor, a memory and a bus connected to the processor; the processor and the memory communicate with each other via the bus; The processor is used to call the computer program in the memory to execute the method for generating early warning service information based on multi-dimensional business monitoring according to any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the method for generating early warning service information based on multi-dimensional business monitoring as described in any one of claims 1-8 is implemented.

Images