Digital enterprise management method and system based on machine learning
Through machine learning-based methods, the digital enterprise log analysis network is identified and optimized, and the problem of unstable log processing performance under different management tasks is solved, and efficient and accurate log analysis is achieved to meet the data processing needs of modern enterprises.
Patent Information
- Application Number
- CN202410545781.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-05-06
- Publication Date
- 2024-11-26
- Estimated Expiration
- 2044-05-06
AI Technical Summary
The existing digital enterprise log analysis network has problems of unstable performance and low accuracy when processing log data under different management task states, and lacks scientific error quantification methods, which leads to inefficient network debugging and difficult to achieve.
Using machine learning-based methods, by obtaining digital enterprise logs under different management tasks, performing text memory semantic recognition, determining the target memory semantic commonality score and semantic thermal correlation between log text memory semantic knowledge, and calculating network debugging errors variables, optimize network bias weights to improve network performance.
It realizes a high-performance and high-stability digital enterprise log analysis network, which can more accurately process and analyze log data under different management task states, meeting the comprehensive needs of modern enterprises for log data processing.
Smart Images

Figure CN118446416B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of digitalization and artificial intelligence technology, and in particular to a digital enterprise management method and system based on machine learning. Background Art
[0002] With the continuous deepening of enterprise informatization, digital enterprise logs have become an indispensable source of data in the enterprise operation process. These logs record detailed information of the enterprise in various management task states, which is of great significance for analyzing the enterprise operation status and optimizing management decisions. However, traditional log analysis methods often rely on manual interpretation, which is inefficient and error-prone, and cannot meet the needs of modern enterprises for efficient, accurate and real-time data processing.
[0003] In order to solve this problem, in recent years, technologies based on digital enterprise log analysis networks have gradually emerged. Such networks can automatically conduct in-depth analysis of enterprise logs and extract valuable information. However, existing log analysis networks often have problems such as unstable performance and low accuracy when processing log data in different management task states. Summary of the invention
[0004] In order to improve the above problems, the present application provides a digital enterprise management method and system based on machine learning.
[0005] The present application embodiment provides a digital enterprise management method based on machine learning, which is applied to a digital enterprise management system. The method includes:
[0006] Acquire a first digital enterprise log of a first management task attention state, enter the first digital enterprise log into a digital enterprise log analysis network to be debugged, and obtain a second digital enterprise log of a second management task attention state;
[0007] Performing text memory semantic recognition on the first digitized enterprise log to obtain first log text memory semantic knowledge; performing text memory semantic recognition on the second digitized enterprise log to obtain second log text memory semantic knowledge;
[0008] Determine a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, obtain a first network debugging error variable according to the target memory semantic commonality score, and the first network debugging error variable has a first quantitative relationship with the target memory semantic commonality score;
[0009] Acquire a first semantic description thermal record corresponding to the first log text memory semantic knowledge and a second semantic description thermal record corresponding to the second log text memory semantic knowledge;
[0010] Determine a semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record, obtain a second network debugging error variable according to the semantic thermal correlation, and the second network debugging error variable has a first quantitative relationship with the semantic thermal correlation;
[0011] Obtaining a target network debugging error variable according to the first network debugging error variable and the second network debugging error variable;
[0012] The network bias weight of the digital enterprise log analysis network is optimized according to the target network debugging error variable to obtain a debugged digital enterprise log analysis network, so as to process the digital enterprise logs using the debugged digital enterprise log analysis network.
[0013] Preferably, the step of obtaining the first semantic description thermal record corresponding to the first log text memory semantic knowledge includes:
[0014] Performing a thermal analysis based on a first semantic description value in the first log text memory semantic knowledge to obtain a first thermal distribution coefficient corresponding to the first semantic description value;
[0015] A first semantic description thermal record corresponding to the first log text memory semantic knowledge is determined based on the first thermal distribution coefficient.
[0016] Preferably, the first thermal distribution coefficient includes a target global indicator and a target discrete indicator, and the determining of the first semantic description thermal record corresponding to the first log text memory semantic knowledge based on the first thermal distribution coefficient includes:
[0017] Acquire a target indication weight set, wherein the target indication weight set includes a plurality of log semantic reconstruction weights;
[0018] Performing an adjustment operation on the target discrete indicator based on the log semantic reconstruction weight to obtain a target semantic adjustment indicator;
[0019] The target global indicator is modified according to the target semantic adjustment indicator to obtain a target log event label value, and the target log event label values corresponding to each log semantic reconstruction weight constitute a target log event label value set;
[0020] Determine the target activity index corresponding to each target log event tag value based on the target log event tag value set, obtain a first semantic description thermal record based on the target activity index corresponding to each target log event tag value, and use the target activity index corresponding to each target log event tag value as the thermal feature in the first semantic description thermal record.
[0021] Preferably, the target activity index is determined based on the global semantic thermal spectrum corresponding to the target log event tag value set; the target indication weight set is obtained, and the target indication weight set includes multiple log semantic reconstruction weights including:
[0022] Acquire semantic thermal reference conditions, extract log event label values according to the semantic thermal reference conditions, and obtain log event label values corresponding to the semantic thermal reference conditions as log semantic reconstruction weights.
[0023] Preferably, the target memory semantic commonality score includes a first semantic decision possibility, and the determining of the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and obtaining a first network debugging error variable according to the target memory semantic commonality score includes:
[0024] combining the first log text memory semantic knowledge and the second log text memory semantic knowledge, and using the combined semantic knowledge as a positive example of memory semantic knowledge;
[0025] Entering the positive examples of the memorized semantic knowledge into a deep semantic decision tree to obtain a first semantic decision possibility;
[0026] A first network debugging error variable is obtained according to the first semantic decision possibility.
[0027] Preferably, obtaining a first network debugging error variable according to the first semantic decision possibility includes:
[0028] Acquire a third digital enterprise log of the attention status of the second management task, wherein the third digital enterprise log is different from the second digital enterprise log;
[0029] Performing text memory semantic recognition on the third digitized enterprise log to obtain text memory semantic knowledge of the third log;
[0030] Taking the target log text memory semantic knowledge and the third log text memory semantic knowledge as negative examples of memory semantic knowledge, entering the negative examples of memory semantic knowledge into a deep semantic decision tree, and obtaining a second semantic decision possibility, wherein the target log text memory semantic knowledge is the first log text memory semantic knowledge or the second log text memory semantic knowledge;
[0031] A first network debugging error variable is obtained according to the first semantic decision possibility and the second semantic decision possibility, and the second semantic decision possibility is in a second quantitative relationship with the first network debugging error variable.
[0032] Preferably, the method further comprises:
[0033] Obtaining a network bias weight optimization vector corresponding to the deep semantic decision tree according to the first network debugging error variable;
[0034] The network bias weight of the deep semantic decision tree is optimized according to the network bias weight optimization vector.
[0035] Preferably, the method further comprises:
[0036] Get the basic digital enterprise log analysis network;
[0037] Debugging the basic digital enterprise log analysis network based on the debugging digital enterprise log to obtain the current digital enterprise log analysis network;
[0038] Entering past digital enterprise logs into the current digital enterprise log analysis network for processing to obtain intermediate digital enterprise logs;
[0039] When it is determined that the target abnormality warning level of the intermediate digitized enterprise log relative to the past digitized enterprise log is greater than the warning level threshold, the process jumps to the step of obtaining the first digitized enterprise log in the first management task attention state.
[0040] Preferably, determining the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record comprises:
[0041] Acquire a first thermal feature from the first semantic description thermal record, and acquire a second thermal feature corresponding to the first thermal feature from the second semantic description thermal record;
[0042] Acquire a characteristic distance between the first thermal characteristic and the second thermal characteristic;
[0043] A semantic thermal correlation is determined based on the feature distance, wherein the semantic thermal correlation has a first quantitative relationship with the feature distance.
[0044] Preferably, the first log text memory semantic knowledge is obtained based on a first NLP semantic embedding algorithm, the second log text memory semantic knowledge is obtained based on a second NLP semantic embedding algorithm, and the step of optimizing the network bias weight of the digital enterprise log analysis network according to the target network debugging error variable to obtain a debugged digital enterprise log analysis network, and using the debugged digital enterprise log analysis network to process the digital enterprise log includes:
[0045] Optimizing the network bias weights of the digital enterprise log analysis network, the first NLP semantic embedding algorithm, and the second NLP semantic embedding algorithm according to the target network debugging error variable to obtain optimized digital enterprise log analysis network, the first NLP semantic embedding algorithm, and the second NLP semantic embedding algorithm;
[0046] Jump to the step of obtaining the first digital enterprise log of the first management task attention state to continue network debugging until the digital enterprise log analysis network is in a stable state, and obtain the debugged digital enterprise log analysis network.
[0047] Preferably, the method further comprises:
[0048] Obtain pending digital enterprise logs for which attention status updates are to be made;
[0049] Determine a first management task attention status matched by the to-be-processed digital enterprise log and a second management task attention status to be updated;
[0050] Determine a debugged digital enterprise log analysis network according to the first management task attention state and the second management task attention state; the digital enterprise log analysis network is debugged based on a target network debugging error variable, the target network debugging error variable is obtained according to the first network debugging error variable and the second network debugging error variable, the first network debugging error variable is determined according to a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, the first network debugging error variable and the target memory semantic commonality score have a first quantitative relationship, the second network debugging error variable is obtained according to a semantic thermal correlation between a first semantic description thermal record corresponding to the first log text memory semantic knowledge and a second semantic description thermal record corresponding to the second log text memory semantic knowledge, the second network debugging error variable and the semantic thermal correlation have a first quantitative relationship, the first log text memory semantic knowledge is a log text memory semantic knowledge corresponding to a first digital enterprise log of the first management task attention state, the second log text memory semantic knowledge is a log text memory semantic knowledge corresponding to a second digital enterprise log of the second management task attention state, and the second digital enterprise log is obtained according to the first digital enterprise log;
[0051] The digital enterprise log to be processed is entered into the debugged digital enterprise log analysis network, and the target digital enterprise log of the second management task focus status is output.
[0052] An embodiment of the present application provides a digital enterprise management system, comprising at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the above method.
[0053] An embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. The computer program implements the above method when running.
[0054] In the embodiment of the present application, the technical problems pointed out in the background technology are mainly because the network fails to fully consider the characteristics and differences of log data under different task states during the design and training process, resulting in the limited processing ability of the network for log data in different states. For example, when the existing log analysis network processes the first digital enterprise log of the first management task focus state, it may be able to extract and analyze the information in the log well, but when processing the log of the second management task focus state, the performance may be greatly reduced. This performance difference between states seriously affects the accuracy and reliability of log analysis. In addition, when the existing log analysis network extracts the text memory semantic knowledge in the log, it often only focuses on a single semantic feature, ignoring the commonality and difference between semantics. This makes it difficult for the network to accurately identify and extract key information when processing log data with similar or related semantics. At the same time, the existing network lacks a scientific and systematic error quantification method in the debugging and optimization process. This makes the network debugging process inefficient and difficult to improve network performance in a targeted manner. In summary, there are many problems and challenges in the existing digital enterprise log analysis network when processing log data under different management task states.
[0055] In order to solve these problems, the embodiment of the present application proposes a new log analysis network debugging and optimization method. By comprehensively considering the characteristics of log data in different states, extracting multi-dimensional semantic knowledge, and adopting scientific error quantification methods, the embodiment of the present application aims to create a high-performance, high-stability digital enterprise log analysis network to meet the comprehensive needs of modern enterprises for log data processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0056] Figure 1 A flowchart of a digital enterprise management method based on machine learning provided in an embodiment of the present application.
[0057] Figure 2 A schematic diagram of the structure of a digital enterprise management system 200 provided in an embodiment of the present application. DETAILED DESCRIPTION
[0058] 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.
[0059] Figure 1 A digital enterprise management method based on machine learning is shown, which is applied to a digital enterprise management system. The method includes the following steps 110 to 170.
[0060] Step 110: Obtain a first digital enterprise log of a first management task attention state, enter the first digital enterprise log into a digital enterprise log analysis network to be debugged, and obtain a second digital enterprise log of a second management task attention state.
[0061] Among them, the first management task attention state refers to a specific attention state set by the digital enterprise management system when conducting digital enterprise management. For example, the first management task attention state focuses on one aspect or a combination of several aspects such as sales data anomalies, user behavior analysis, and system performance monitoring. In this state, the digital enterprise management system can collect relevant information for subsequent analysis and decision support. The first digital enterprise log refers to the log file of relevant operations, events or data recorded by the digital enterprise management system in the first management task attention state. The first digital enterprise log contains user behavior data, system operation status, transaction records, exception reports, etc., and is an important data source for understanding and analyzing the operation of the enterprise.
[0062] Furthermore, the digital enterprise log analysis network to be debugged is a neural network model that has not been optimized or debugged yet, and is specifically used to process log files of digital enterprises. The digital enterprise log analysis network to be debugged needs to continuously learn and adjust to improve its accuracy and efficiency in processing and analyzing log data.
[0063] Correspondingly, the second management task focus state is a state related to another management focus point output after the first digital enterprise log is processed by the digital enterprise log analysis network to be debugged. The second management task focus state can be set according to specific business needs or analysis goals for further data mining or decision support. The second digital enterprise log is log data processed and output by the digital enterprise log analysis network in the second management task focus state. The second digital enterprise log has undergone preliminary analysis and processing by the digital enterprise log analysis network, and contains more refined and targeted information, which is convenient for managers or analysts to further interpret and apply data.
[0064] In detail, step 110 involves clarifying the current management focus: the first management task focus state. The first management task focus state reflects the business level that the digital enterprise management system wants to deeply understand or monitor, and can focus on the fluctuation of sales data, user behavior patterns, or system performance. Based on this focus, the relevant log files carefully selected from the digital enterprise system can be used as the first digital enterprise log.
[0065] The first digital enterprise log records in detail various operations, events and data exchanges under a specific state of concern, and is an information source for understanding the operation status of the enterprise. Next, the first digital enterprise log is entered into the digital enterprise log analysis network to be debugged. The digital enterprise log analysis network to be debugged is a powerful data processing tool that can conduct in-depth analysis and mining of a large amount of log data (such as the first digital enterprise log).
[0066] After the first digital enterprise log is processed by the digital enterprise log analysis network to be debugged, data focusing on another management concern is obtained: the second digital enterprise log. The second digital enterprise log has been preliminarily screened and analyzed by the digital enterprise log analysis network to be debugged, and is more refined and targeted, revealing the enterprise operation status from different management perspectives. In this way, step 110 not only completes the collection and preliminary processing of data, but also lays a solid foundation for subsequent in-depth data interpretation and application.
[0067] Step 120: Perform text memory semantic recognition on the first digitized enterprise log to obtain first log text memory semantic knowledge; perform text memory semantic recognition on the second digitized enterprise log to obtain second log text memory semantic knowledge.
[0068] Among them, text memory semantic recognition is a natural language processing technology, that is, by analyzing the semantic content of the text, extracting and understanding the meaning and contextual relationship in the text. Text memory semantic recognition not only focuses on the vocabulary and grammar of the text, but also digs deep into the deep concepts, emotions and intentions expressed by the text. In text memory semantic recognition, the relevance, temporal sequence and historical interaction information between texts are also considered to more comprehensively understand the text content.
[0069] Correspondingly, the first log text memory semantic knowledge refers to the valuable semantic information and knowledge extracted from the first digital enterprise log through text memory semantic recognition technology. The first log text memory semantic knowledge reflects the core content and intention in the first digital enterprise log, which can include user behavior patterns, system operation processes, transaction activity characteristics, etc. By analyzing the first log text memory semantic knowledge, its operating status and user needs can be better understood. Similar to the first log text memory semantic knowledge, the second log text memory semantic knowledge is the semantic information and knowledge extracted from the second digital enterprise log through text memory semantic recognition technology. Since the second digital enterprise log is obtained under different management task attention states, the second log text memory semantic knowledge reveals different business levels and user behavior characteristics, thereby providing a more comprehensive operational perspective.
[0070] It can be seen that step 120 involves in-depth understanding and knowledge extraction of the first and second digital enterprise logs. First, text memory semantic recognition is performed on the first digital enterprise log. Text memory semantic recognition is not just a simple text analysis, but uses natural language processing technology to deeply explore the semantic information and contextual relationships in the log. Through text memory semantic recognition, the core content and intention in the log can be captured, and key features such as user behavior, system operation, and transaction activities can be understood. In this way, the first log text memory semantic knowledge is obtained, and these first log text memory semantic knowledge presents the operation of the enterprise under the first management task focus state in a structured feature form. Next, the second digital enterprise log is subjected to the same text memory semantic recognition process. Since the second digital enterprise log focuses on different management concerns, the semantic knowledge and information extracted through recognition will also be different. Through the second log text memory semantic knowledge, the operation status of the enterprise under different business levels and user behavior characteristics can be more comprehensively understood. Step 120 not only completes the in-depth understanding of the digital enterprise log, but also provides a rich semantic knowledge and information foundation for subsequent data analysis, business optimization and decision support. By comparing and analyzing the semantic knowledge of the first log text memory with the semantic knowledge of the second log text memory, it is possible to gain insights into potential problems and opportunities in their operations, thereby making more informed decisions.
[0071] Step 130: determine a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and obtain a first network debugging error variable based on the target memory semantic commonality score. The first network debugging error variable has a first quantitative relationship with the target memory semantic commonality score.
[0072] Among them, the target memory semantic commonality score is a scoring indicator used to measure the commonality between the first log text memory semantic knowledge and the second log text memory semantic knowledge. The target memory semantic commonality score reflects the similarity and consistency between the first log text memory semantic knowledge and the second log text memory semantic knowledge in terms of semantic content, contextual relationship, and intention expression. The higher the score, the more commonality there is between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and vice versa, it means that the first log text memory semantic knowledge and the second log text memory semantic knowledge are more different.
[0073] The first network debugging error variable (first debugging loss) is a variable used to represent the gap between the actual output and the expected output during the network debugging process. In an embodiment of the present application, the first network debugging error variable is calculated based on the target memory semantic commonality score, which is used to guide the further debugging and optimization of the network. For example, when the target memory semantic commonality score is low, it means that there is a difference in the network when processing the two log text memory semantic knowledge, and it is necessary to reduce this difference by adjusting the network parameters.
[0074] Furthermore, the first quantitative relationship refers to a quantifiable mathematical relationship between the first network debugging error variable and the target memory semantic commonality score. For example, if the first network debugging error variable and the target memory semantic commonality score show a negative correlation, then when the target memory semantic commonality score decreases, the first network debugging error variable will increase accordingly, and vice versa. This quantitative relationship helps to more accurately understand the relationship between network performance and network output quality.
[0075] It can be understood that step 130 involves the evaluation of the commonality between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and the calculation of the network debugging error based on this commonality. First, it is necessary to determine the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge. The target memory semantic commonality score is obtained by comparing the similarity and consistency of the two parts of knowledge in terms of semantics, context and intention. The level of the target memory semantic commonality score directly reflects the degree of matching between the first log text memory semantic knowledge and the second log text memory semantic knowledge. Next, the first network debugging error variable is calculated based on the target memory semantic commonality score. The first network debugging error variable actually represents the difference or deviation in the network when processing the two log text memory semantic knowledge. If the target memory semantic commonality score is high, it means that the network can better understand and process the two log text memory semantic knowledge, and the first network debugging error variable will be relatively small at this time. On the contrary, if the target memory semantic commonality score is high, it means that there is a large deviation in the network during the processing process, and further debugging is needed to optimize the network performance. In addition, it is also necessary to note the first quantitative relationship between the first network debugging error variable and the target memory semantic commonality score. The first quantitative relationship can be negatively correlated, that is, when the commonality score decreases, the error variable will increase. The first quantitative relationship is of great significance for understanding and optimizing network performance because it can help locate the problem more accurately, thereby adjusting the network parameters in a targeted manner to improve the processing effect. Based on this, step 130 provides an effective means of evaluating and optimizing network performance by calculating the target memory semantic commonality score and the first network debugging error variable.
[0076] Step 140: Obtain a first semantic description thermal record corresponding to the first log text memory semantic knowledge and a second semantic description thermal record corresponding to the second log text memory semantic knowledge.
[0077] Among them, the first semantic description heat record is used to intuitively display the importance and distribution of each semantic element in the semantic knowledge of the first log text memory. The first semantic description heat record can represent the frequency of occurrence, attention or importance of different semantic elements through visual elements such as color, size, and shape. For example, in a heat map, high-frequency or important semantic elements may be given brighter or darker colors and a larger display size to visually highlight their importance. In this way, the first semantic description heat record can help analysts quickly identify and understand key information and trends in the semantic knowledge of the first log text memory.
[0078] Similar to the first semantic description heat record, the second semantic description heat record is a visual display of the semantic knowledge memorized in the second log text. The second semantic description heat record reflects the importance and distribution of each semantic element in the second log. By comparing the first and second semantic description heat records, we can gain insight into the similarities and differences between the semantic knowledge memorized in the first and second log texts, thereby gaining a more comprehensive understanding of the enterprise's operating status and user needs.
[0079] Therefore, step 140 involves obtaining and displaying the heat record of the first log text memory semantic knowledge and the second log text memory semantic knowledge. First, it is necessary to extract key semantic elements from the first log text memory semantic knowledge that has been identified by text memory semantics, and quantitatively evaluate them according to their frequency of occurrence, attention or importance. These key semantic elements may include specific words, phrases or concepts, which appear repeatedly in the log and are of great significance for understanding the operating status and user needs of the enterprise. Next, using visualization tools such as heat maps, these key semantic elements are displayed in an intuitive way to form a first semantic description heat record. In the first semantic description heat record, each semantic element is given a specific visual identifier, such as color, size or shape, to reflect its importance and distribution in the first log text memory semantic knowledge. Similarly, the same operation is performed on the second log text memory semantic knowledge to extract key semantic elements and generate a second semantic description heat record. Since the second digitized enterprise log focuses on different management concerns, its heat record may show different characteristics and trends from the first digitized enterprise log. By acquiring and comparing these two thermal records, we can gain a deeper understanding of the enterprise's operations and user needs under different management task focus states. For example, some semantic elements may occupy an important position in both thermal records, indicating that these elements are the core focus of the enterprise's operations; while some elements may only be prominent in one thermal record, which helps to discover specific needs and challenges under different management tasks. In summary, step 140 provides an intuitive and comprehensive perspective to analyze and understand the enterprise's log text memory semantic knowledge by acquiring and displaying the first and second semantic description thermal records.
[0080] Step 150: determine the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record, obtain a second network debugging error variable based on the semantic thermal correlation, and the second network debugging error variable has a first quantitative relationship with the semantic thermal correlation.
[0081] Among them, semantic thermal correlation refers to the correlation between the first semantic description thermal record and the second semantic description thermal record in terms of semantic content and thermal distribution. For example, semantic thermal correlation measures the similarity and consistency of the two thermal records in terms of the occurrence, frequency, and importance of key semantic elements. The level of semantic thermal correlation can reflect the degree of association and complementarity between the semantic knowledge remembered in the two log texts.
[0082] The second network debugging error variable is an error variable calculated based on the semantic thermal correlation during the network debugging process. The second network debugging error variable represents the difference or deviation when the network processes two semantically described thermal records, and is used to guide further debugging and optimization of the network. When the semantic thermal correlation is low, it means that there are differences in the network's understanding and processing of the semantic content and thermal distribution of the two thermal records, and this difference needs to be reduced by adjusting the network parameters.
[0083] In detail, step 150 involves the evaluation of the correlation between the first semantic description thermal record and the second semantic description thermal record, and the calculation of the network debugging error based on this correlation. First, it is necessary to determine the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record. The semantic thermal correlation is measured by comparing the similarity and consistency of the two thermal records in terms of key semantic elements, frequency of occurrence, importance, etc. The level of semantic thermal correlation reflects the degree of matching between the two thermal records in terms of semantic content and thermal distribution. Next, the second network debugging error variable is calculated based on this semantic thermal correlation. The second network debugging error variable actually represents the difference or deviation in the network when processing the two thermal records. If the semantic thermal correlation is high, it means that the network can better understand and process the semantic content and thermal distribution of the two thermal records, and the second network debugging error variable will be relatively small at this time. On the contrary, if the semantic thermal correlation is low, it means that there is a large deviation in the network during the processing process, and further debugging is needed to optimize the network's understanding and processing capabilities of the two semantic description thermal records. Similar to step 130, it is also necessary to note the first quantitative relationship between the second network debugging error variable and the semantic thermal correlation. This relationship can be negatively correlated, that is, when the correlation decreases, the error variable increases. This quantitative relationship is of great significance for understanding and optimizing the performance of the network in understanding and processing two thermal records, because it can help locate the problem more accurately, so as to adjust the network parameters in a targeted manner to improve the processing effect. It can be seen that step 150 provides an effective means of evaluating and optimizing the performance of the network in understanding and processing two thermal records by calculating the semantic thermal correlation and the second network debugging error variable, which helps to more comprehensively understand the performance of the network under different management task attention states, thereby providing enterprises with more accurate and valuable data analysis and decision support.
[0084] Step 160: Obtain a target network debugging error variable according to the first network debugging error variable and the second network debugging error variable.
[0085] Among them, the target network debugging error variable is a comprehensive error variable, which is obtained by comprehensively considering and calculating the first network debugging error variable and the second network debugging error variable. The target network debugging error variable not only reflects the performance difference of the network in processing the semantic knowledge of the first log text memory and the semantic knowledge of the second log text memory, but also reflects the network's ability to understand and process the semantic thermal records of the first and second digital enterprise logs. The target network debugging error variable is a key indicator in the process of network debugging and optimization. Through the target network debugging error variable, the performance of the network can be evaluated more comprehensively, and more accurate network parameter adjustments can be made.
[0086] Correspondingly, the core goal of step 160 is to obtain a comprehensive target network debugging error variable based on the first network debugging error variable and the second network debugging error variable. When executing this step, it is first necessary to review and analyze the first network debugging error variable and the second network debugging error variable calculated in the previous step. The first network debugging error variable and the second network debugging error variable respectively reflect the performance difference of the network in processing two different log text memory semantic knowledge and their corresponding semantic thermal records. Next, it is necessary to design a reasonable algorithm or model to comprehensively consider the two error variables and calculate a comprehensive target network debugging error variable. This algorithm or model can be a weighted average model, a fusion model based on machine learning, or a more complex optimization algorithm, depending on the actual application scenario and requirements. In the process of calculating the target network debugging error variable, the weight of the two error variables may also need to be considered. For example, if it is believed that the ability to process log text memory semantic knowledge is more important than the ability to process semantic thermal records, then the first network debugging error variable can be given a greater weight. Finally, through the calculated target network debugging error variable, the overall performance of the network can be more comprehensively evaluated, and the problems and deficiencies of the network in processing two log texts and their thermal records can be found. This comprehensive target network debugging error variable will provide a clear and quantitative indicator to guide the subsequent network debugging and optimization work. Therefore, step 160 is a comprehensive evaluation and optimization step, which obtains a more comprehensive and accurate target network debugging error variable by comprehensively considering and processing the two network debugging error variables obtained in the previous step, providing strong support and guidance for the subsequent network debugging and optimization work.
[0087] Step 170: Optimize the network bias weights of the digital enterprise log analysis network according to the target network debugging error variable to obtain a debugged digital enterprise log analysis network, so as to process digital enterprise logs using the debugged digital enterprise log analysis network.
[0088] Among them, the network bias weight is an important parameter for adjusting the output of neurons, which is similar to adding an adjustable constant term to each neuron, so that the output of the neuron can be fine-tuned without relying entirely on the input data. Network bias weights play a vital role in network learning because they can help the network better fit the data and improve the accuracy and generalization of predictions. In the digital enterprise log analysis network, optimizing the network bias weight can further improve the network's processing capabilities and analysis accuracy of log data (digital enterprise logs).
[0089] Accordingly, the debugged digital enterprise log analysis network is an optimized and adjusted neural network model specifically designed for processing and analyzing log data of digital enterprises. By optimizing parameters such as network bias weights, the debugged digital enterprise log analysis network can more accurately identify and extract key information from logs, providing valuable insights and analysis results for enterprises. The debugged digital enterprise log analysis network is an important tool for enterprise data analysis and decision support. It can help enterprises better understand user behavior, optimize product services, and improve operational efficiency.
[0090] It can be seen that step 170 is a key link in the embodiment of the present application, which involves optimizing the network bias weight of the digital enterprise log analysis network according to the target network debugging error variable, and finally obtaining the debugged digital enterprise log analysis network. The following is an exemplary introduction to this step.
[0091] 1. Preparation
[0092] Before starting the optimization, you need to ensure that the target network debugging error variable has been calculated and the optimization goal and direction have been clarified.
[0093] At the same time, prepare the infrastructure and initial parameter settings of the digital enterprise log analysis network, including the number of network layers, number of neurons, activation functions, etc.
[0094] 2. Optimizing Network Bias Weights
[0095] Analyze the target network debugging error variables: First, carefully analyze the target network debugging error variables to understand the source and size of the network error when processing log data. This will help determine the specific parameters and directions that need to be optimized.
[0096] Adjust network bias weights: Based on the results of the error analysis, begin adjusting the bias weights of the digital enterprise log analysis network. This typically involves using an optimization algorithm (such as gradient descent or its variants) to gradually adjust the weight values to reduce the value of the target network debugging error variable.
[0097] Iterative optimization process: The optimization process is often iterative, that is, it is necessary to adjust the bias weights and evaluate the performance of the network multiple times until a satisfactory performance level is achieved or the error is reduced to an acceptable range.
[0098] 3. Verification and Testing
[0099] During the optimization process, a validation dataset is used to evaluate the performance of the network. This helps ensure that the network not only performs well on the training data, but also maintains good generalization capabilities on unseen data.
[0100] Once the network bias weights are optimized, a test dataset can be used to further validate the performance of the network and ensure that it meets actual needs.
[0101] 4. Get a debugged digital enterprise log analysis network
[0102] After the above optimization and verification process, a debugged digital enterprise log analysis network was obtained. This network has been optimized and adjusted to better process and analyze the log data of the digital enterprise.
[0103] 5. Apply the debugged network
[0104] Finally, the debugged digital enterprise log analysis network can be used to process and analyze the actual digital enterprise logs. This will provide valuable insights and decision support for enterprises, helping them better understand user behavior, optimize product services, and improve operational efficiency.
[0105] Based on the above content, the following is an introduction through a complete application scenario.
[0106] In the daily operation process, the digital enterprise management system will continuously collect and record various log data of the enterprise. In order to analyze this data more effectively, the system decided to debug and optimize its built-in log analysis network.
[0107] First, the digital enterprise management system obtains the first digital enterprise log in the state of attention of a specific first management task. The first digital enterprise log records the enterprise operation data in detail in the task state. Subsequently, the system enters the first digital enterprise log into the digital enterprise log analysis network to be debugged, and obtains the second digital enterprise log in the state of attention of the second management task.
[0108] Next, the system performs text memory semantic recognition on the first digitized enterprise log, extracts key information and semantic features, and forms the first log text memory semantic knowledge. Similarly, the system performs the same processing on the second digitized enterprise log and obtains the second log text memory semantic knowledge.
[0109] In order to evaluate the performance of the log analysis network in processing log data (digital enterprise logs) under different management task states, the digital enterprise management system determines the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge. This score reflects the similarities and differences between the two log text memory semantic knowledge. Based on the target memory semantic commonality score, the system calculates the first network debugging error variable, which has a clear quantitative relationship with the target memory semantic commonality score.
[0110] In addition, the system also obtains the thermal record of the first semantic description corresponding to the first log text memory semantic knowledge and the thermal record of the second semantic description corresponding to the second log text memory semantic knowledge. These two thermal records show the distribution and focus of log text memory semantic knowledge in an intuitive way.
[0111] By comparing the first semantic description thermal record with the second semantic description thermal record, the digital enterprise management system determines the semantic thermal correlation between them. This correlation index helps the system understand the network's recognition and emphasis on key semantic elements when processing log data in different management task states. Based on this semantic thermal correlation, the system further calculates the second network debugging error variable, which also has a clear quantitative relationship with the semantic thermal correlation.
[0112] After obtaining the first network debugging error variable and the second network debugging error variable, the digital enterprise management system comprehensively calculates the target network debugging error variable based on these two variables. The target network debugging error variable comprehensively reflects the performance differences and deficiencies of the log analysis network when processing log data under different management task states.
[0113] Finally, the digital enterprise management system optimizes the network bias weight of the digital enterprise log analysis network based on the target network debugging error variable. By adjusting the network parameters and configuration, the system obtains a debugged digital enterprise log analysis network. This optimized network can more accurately analyze and process the enterprise's log data under different management task states, providing enterprises with more valuable operational insights and decision support.
[0114] By applying the embodiment of the present application, firstly, by obtaining the digital enterprise logs under different management task attention states and entering them into the digital enterprise log analysis network to be debugged, the network's processing ability for log data in different task states can be comprehensively evaluated, which not only ensures the diversity and practicality of log data, but also provides a solid foundation for subsequent network debugging and optimization. Secondly, the text memory semantic recognition technology is used to conduct in-depth analysis of the log data, extract key text memory semantic knowledge, enhance the depth and accuracy of log analysis, and enable the network to better understand and process complex information in the enterprise log. In addition, by determining the target memory semantic commonality score between different log text memory semantic knowledge, the performance difference of the network in processing similar semantic information can be accurately quantified, thereby obtaining the first network debugging error variable. This quantification method makes the network debugging process more scientific and systematic, and helps to improve network performance in a targeted manner. At the same time, the analysis of semantic description thermal records is innovatively introduced, and the second network debugging error variable is further obtained by determining the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record. This step not only takes into account the commonality of semantics, but also pays attention to the thermal distribution of semantics, providing strong support for the comprehensive optimization of the network. Finally, based on the comprehensive target network debugging error variable, the network bias weight of the digital enterprise log analysis network can be accurately optimized, thereby obtaining a debugged network model with better performance. This optimization process not only improves the network's processing efficiency and accuracy of log data, but also provides enterprises with more reliable and efficient data analysis tools, helping them make more informed decisions.
[0115] In some optional embodiments, obtaining the first semantic description thermal record corresponding to the first log text memory semantic knowledge includes: performing a thermal analysis based on the first semantic description value in the first log text memory semantic knowledge to obtain a first thermal distribution coefficient corresponding to the first semantic description value; and determining the first semantic description thermal record corresponding to the first log text memory semantic knowledge based on the first thermal distribution coefficient.
[0116] In this embodiment, a key step in the digital enterprise management system during the process of performing log analysis network debugging and optimization is to obtain the semantic description thermal records corresponding to the log text memory semantic knowledge. This step is crucial for evaluating the network's understanding and processing capabilities of log data.
[0117] Exemplarily, after obtaining the first digital enterprise log in the first management task attention state, the digital enterprise management system first performs text memory semantic recognition on the first digital enterprise log, thereby obtaining the first log text memory semantic knowledge. The first log text memory semantic knowledge reflects the key information and semantic features in the log, and is the basis for subsequent analysis.
[0118] Next, the system can perform thermal analysis to more intuitively understand the distribution of these semantic knowledge in the log. In thermal analysis, the digital enterprise management system will first focus on the first semantic description value in the first log text memory semantic knowledge. The first semantic description value can be some keywords, phrases or concepts, and their frequency and position in the log will affect its thermal distribution.
[0119] The system can conduct in-depth analysis of these first semantic description values and calculate the thermal distribution coefficient of each description value in the log. The thermal distribution coefficient actually reflects the importance and distribution of the description value in the log. For example, if a keyword appears frequently in the log and is in a critical position, its thermal distribution coefficient will be relatively high.
[0120] Based on these first thermal distribution coefficients, the digital enterprise management system can determine the first semantic description thermal record corresponding to the first log text memory semantic knowledge. The first semantic description thermal record is like a "heat map", which clearly shows which semantic descriptions are "hot" in the log and which are relatively "unpopular".
[0121] Similarly, for the second digital enterprise log processed by the log analysis network, the system will also perform a similar thermal analysis to obtain a second semantic description thermal record corresponding to the semantic knowledge memorized in the second log text.
[0122] By comparing the thermal records of the first semantic description with the thermal records of the second semantic description, the digital enterprise management system can analyze the network's ability to capture key semantic descriptions when processing logs. If the thermal distribution of the two is similar, it means that the network has better preserved the key information in the original log; if the difference is significant, it may mean that the network has lost or distorted some important information during the processing.
[0123] In summary, through the thermal analysis method, the digital enterprise management system can have a deeper understanding of the log analysis network's ability to process log data under different management task states, and provide strong data support for subsequent network debugging and optimization.
[0124] In this way, by introducing thermal analysis technology, the digital enterprise management system can more intuitively evaluate and optimize the performance of the log analysis network. For example, by comparing the semantic description thermal records in different states, the system can not only find the deficiencies in the network processing logs, but also debug and optimize them in a targeted manner. This method based on thermal analysis improves the accuracy and reliability of log analysis, and also enhances the digital enterprise management system's processing capabilities and insight into log data.
[0125] In some other optional embodiments, the first thermal distribution coefficient includes a target global indicator and a target discrete indicator, and determining the first semantic description thermal record corresponding to the first log text memory semantic knowledge based on the first thermal distribution coefficient includes: obtaining a target indication weight set, the target indication weight set including multiple log semantic reconstruction weights; adjusting the target discrete indicator based on the log semantic reconstruction weight to obtain a target semantic adjustment indicator; modifying the target global indicator according to the target semantic adjustment indicator to obtain a target log event label value, and the target log event label values corresponding to each log semantic reconstruction weight constitute a target log event label value set; determining the target activity index corresponding to each target log event label value according to the target log event label value set, obtaining the first semantic description thermal record according to the target activity index corresponding to each target log event label value, and using the target activity index corresponding to each target log event label value as the thermal feature in the first semantic description thermal record.
[0126] In digital enterprise management systems, in-depth analysis and processing of log data is a complex and sophisticated process. Especially when conducting thermal analysis, the system needs to consider multiple dimensions and indicators to ensure the accuracy and comprehensiveness of the analysis results.
[0127] In this embodiment, after the system obtains the first log text memory semantic knowledge, it will further calculate the corresponding first thermal distribution coefficient. The first thermal distribution coefficient here is not just a single value, it actually includes two important components: the target global index and the target discrete index.
[0128] The target global index is an overall evaluation of the semantic knowledge of the entire log text memory, which reflects the distribution and importance of semantic knowledge in the global scope. The target discrete index focuses more on local details, which can help the system identify those discrete semantic points that may contain important information although they account for a small proportion of the entire log.
[0129] In order to obtain a more accurate thermal distribution coefficient, the digital enterprise management system can first obtain a target indication weight set. This weight set contains multiple log semantic reconstruction weights, each of which corresponds to the importance of a specific semantic or event in the log. These weights are usually set based on historical data, expert knowledge, or the output of a machine learning algorithm.
[0130] Next, the system can use these log semantic reconstruction weights to adjust the target discrete indicators. The purpose of this step is to balance global and local information and ensure that semantic points that are ignored by global indicators but are critical in certain specific situations can be correctly identified and weighted.
[0131] The adjusted target discrete index will be further used to modify the target global index, thus obtaining a series of target log event label values. These label values not only take into account the global information in the log, but also incorporate key local details. Each log semantic reconstruction weight will correspond to a target log event label value, and these label values together constitute the target log event label value set.
[0132] Based on the target log event tag value set, the digital enterprise management system can further determine the target activity index corresponding to each target log event tag value. The target activity index actually reflects the activity or importance of a log event or semantic point under the attention of a specific management task.
[0133] Finally, the system can generate the first semantic description thermal record based on these target activity indexes. This record not only contains the global thermal distribution, but also highlights those key local hot spots. By using the target activity index corresponding to each target log event tag value as a thermal feature, this record provides strong data support for subsequent log analysis, network debugging and optimization.
[0134] In this way, the global and local thermal analysis techniques can be used in combination to deeply process log data, which not only improves the accuracy of log analysis, but also enables the system to more comprehensively understand the key information and semantic features in the log. In particular, by introducing the concepts of target indication weight set and activity index, the system can more accurately identify and weight those semantic points that are critical in different management task attention states. This not only enhances the depth and breadth of log analysis, but also provides enterprises with more reliable data support, helping enterprises make more informed and accurate decisions.
[0135] In some preferred embodiments, the target activity index is determined based on the global semantic thermal spectrum corresponding to the target log event label value set; the target indication weight set is obtained, and the target indication weight set includes multiple log semantic reconstruction weights, including: obtaining semantic thermal reference conditions, extracting log event label values based on the semantic thermal reference conditions, and obtaining log event label values corresponding to the semantic thermal reference conditions as log semantic reconstruction weights.
[0136] In actual application, the calculation and determination of the target activity index is particularly critical. The target activity index not only reflects the activity level of each tag in the log event, but also serves as an important basis for system optimization and decision-making. The determination of the target activity index is not made out of thin air, but is determined based on the global semantic thermal spectrum corresponding to the target log event tag value set.
[0137] In short, the global semantic thermal spectrum is a description of the global thermal distribution of the entire log event tag value set. The global semantic thermal spectrum can help the system understand from a macro perspective which log event tags are more active globally and which are relatively quiet. By analyzing and comparing the global semantic thermal spectrum, the digital enterprise management system can more accurately assign a corresponding activity index to each log event tag value.
[0138] Before calculating the target activity index, the system first needs to obtain a key data set: the target indication weight set. This target indication weight set is not generated randomly, but is obtained through a series of precise calculations and extractions. It contains multiple log semantic reconstruction weights, each of which corresponds to a specific log event or semantics.
[0139] In the process of obtaining this target indication weight set, the system will first set a semantic thermal reference condition. The semantic thermal reference condition can be based on the statistical results of historical data or set by experts based on actual needs. In any case, the semantic thermal reference condition is the key to ensure that the subsequently extracted log event label values are representative and accurate.
[0140] With this semantic thermal reference condition, the digital enterprise management system will begin to extract the log event tag values. The system can traverse the entire log database and find the log event tag values that meet the semantic thermal reference conditions. These extracted tag values not only reflect the key information and semantic features in the log, but also represent the activity level of each log event under specific conditions.
[0141] These extracted log event tag values will be used by the system as log semantic reconstruction weights and added to the target indication weight set. In this way, the system obtains a target indication weight set that includes multiple important log semantic reconstruction weights. The target indication weight set not only provides a data basis for the subsequent calculation of the target activity index, but also provides strong support for system optimization and decision-making.
[0142] In this way, through precise calculation and extraction, a target indication weight set containing multiple important log semantic reconstruction weights can be obtained, and the target activity index of each log event tag value can be determined based on the global semantic thermal spectrum. This not only improves the accuracy and efficiency of log analysis, but also provides strong data support for enterprise decision-making and optimization. Through global thermal analysis and precise extraction of weights, the system can more comprehensively understand the key information and semantic features in the log, so as to make more intelligent and accurate decisions.
[0143] In other preferred embodiments, the target memory semantic commonality score includes a first semantic decision possibility, and the determining of the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and obtaining a first network debugging error variable based on the target memory semantic commonality score includes: combining the first log text memory semantic knowledge and the second log text memory semantic knowledge, and using the combined semantic knowledge as a positive example of memory semantic knowledge; entering the positive example of memory semantic knowledge into a deep semantic decision tree to obtain a first semantic decision possibility; and obtaining a first network debugging error variable based on the first semantic decision possibility.
[0144] In practical applications, the commonality between the memory semantic knowledge of two log texts can be evaluated through a new scoring mechanism, namely the target memory semantic commonality score. This score not only deeply reflects the similarity between the two, but also can more accurately locate the possible errors in the network debugging process, thereby optimizing the system performance.
[0145] In the process of determining the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, the digital enterprise management system will first organically integrate the two memory semantic knowledge. This integration is by no means a simple merger, but through a specific algorithm, the core information and features of the two memory semantic knowledge are refined and synthesized into a more comprehensive and representative memory semantic knowledge, which is defined as a positive example of memory semantic knowledge.
[0146] Next, the digital enterprise management system can input the positive examples of memory semantic knowledge into an advanced machine learning model: deep semantic decision tree. The deep semantic decision tree can deeply analyze the intrinsic characteristics and attributes of the input data and output a specific probability value, called the first semantic decision possibility. The first semantic decision possibility is a specific probability value, which indicates the probability of commonality between two pieces of memory semantic knowledge under given conditions.
[0147] After obtaining the first semantic decision possibility, the digital enterprise management system can calculate the first network debugging error variable based on this probability value. The first network debugging error variable can directly reflect the accuracy of the system's evaluation of the commonality of the semantic knowledge of the two log text memories under the current network configuration and parameter settings. If the error variable value is large, it indicates that there is a certain deviation in the system's evaluation result, and corresponding network debugging and system optimization are required.
[0148] If the value of the first semantic decision possibility is high, it means that the commonality between the two pieces of memory semantic knowledge is strong, and the error of network debugging may be relatively small; on the contrary, if the value of the first semantic decision possibility is low, it means that the commonality between the two pieces of memory semantic knowledge is weak, and the error that may exist in the network debugging process will also be relatively large. Based on this error variable, the digital enterprise management system can make targeted adjustments to network parameters and configurations to improve the accuracy of the evaluation of the commonality of log text memory semantic knowledge.
[0149] This design can use the target memory semantic commonality score to comprehensively evaluate the commonality between the semantic knowledge of the two log text memories, and accurately calculate the first network debugging error variable based on the first semantic decision possibility. In this way, not only the accuracy of the system's evaluation of the commonality of semantic knowledge is significantly improved, but also solid data support is provided for subsequent network debugging and system optimization. Through the in-depth application of the deep semantic decision tree model, the digital enterprise management system can more intelligently analyze the deep semantic features in the log text, thereby making more accurate and efficient decisions and adjustments, which is of immeasurable importance to comprehensively improving the comprehensive performance and operating efficiency of the digital enterprise management system.
[0150] In the next step, obtaining the first network debugging error variable based on the first semantic decision possibility includes: obtaining a third digitized enterprise log of the second management task attention status, wherein the third digitized enterprise log is different from the second digitized enterprise log; performing text memory semantic recognition on the third digitized enterprise log to obtain third log text memory semantic knowledge; taking the target log text memory semantic knowledge and the third log text memory semantic knowledge as negative examples of memory semantic knowledge, entering the negative examples of memory semantic knowledge into a deep semantic decision tree to obtain a second semantic decision possibility, wherein the target log text memory semantic knowledge is the first log text memory semantic knowledge or the second log text memory semantic knowledge; obtaining the first network debugging error variable based on the first semantic decision possibility and the second semantic decision possibility, wherein the second semantic decision possibility is in a second quantitative relationship with the first network debugging error variable.
[0151] In the next step, the digital enterprise management system will obtain the first network debugging error variable according to the first semantic decision possibility. This process involves multiple key steps, which will be explained in detail below.
[0152] First, the digital enterprise management system can obtain the third digital enterprise log of the second management task focus state. The third digital enterprise log mentioned here is different from the second digital enterprise log obtained previously, and they respectively record the operation status of the enterprise at different times or different task states.
[0153] After obtaining the third-party digital enterprise log, the system can perform text memory semantic recognition on it. This step is to deeply analyze and understand the text content in the log through specific algorithms and technologies, so as to extract key semantic information. After this step, the system can obtain the third-party log text memory semantic knowledge, which is a highly abstract and generalized log content.
[0154] Next, the digital enterprise management system can combine the target log text memory semantic knowledge with the third log text memory semantic knowledge to form a negative example of memory semantic knowledge. The target log text memory semantic knowledge here can be the first log text memory semantic knowledge or the second log text memory semantic knowledge that has been acquired and processed before. Combining these two parts of knowledge together can form a more comprehensive and rich semantic knowledge base for subsequent analysis and decision-making.
[0155] The system can then feed this memorized negative example of semantic knowledge into a deep semantic decision tree. A deep semantic decision tree is a powerful machine learning model that can automatically make decisions and classifications based on input semantic knowledge. By inputting negative examples, the model can learn which semantic features are not conducive to specific management tasks, thereby avoiding similar mistakes in future decisions.
[0156] After the deep semantic decision tree processes the negative examples, it outputs a second semantic decision possibility. This possibility is also a probability value, which indicates the probability that the model makes the correct decision given the negative examples. This probability value is crucial for evaluating the performance of the model and adjusting the network parameters.
[0157] Finally, the digital enterprise management system can calculate the first network debugging error variable based on the first semantic decision possibility and the second semantic decision possibility. The first network debugging error variable reflects the performance difference of the model when processing positive examples and negative examples, and is an important basis for optimizing the model and network configuration. There is a second quantitative relationship between the second semantic decision possibility and the first network debugging error variable, that is, as the second semantic decision possibility changes, the first network debugging error variable will also be adjusted accordingly.
[0158] In this way, the first semantic decision possibility and the second semantic decision possibility can be fully utilized to accurately calculate the first network debugging error variable. This method not only improves the accuracy of model performance evaluation, but also provides strong data support for network debugging and optimization. By continuously iterating and adjusting model parameters, the system can gradually reduce the first network debugging error variable, thereby improving the accuracy and efficiency of the overall decision.
[0159] In some other possible embodiments, the method further includes: obtaining a network bias weight optimization vector corresponding to the deep semantic decision tree based on the first network debugging error variable; and optimizing the network bias weight of the deep semantic decision tree based on the network bias weight optimization vector.
[0160] In some possible embodiments of the digital enterprise management system, the method further includes the step of adjusting and optimizing the deep semantic decision tree according to the first network debugging error variable. This process is a key link in the self-learning and improvement of the system, and is intended to improve the decision accuracy and generalization ability of the decision tree.
[0161] First, the digital enterprise management system can calculate the network bias weight optimization vector corresponding to the deep semantic decision tree based on the first network debugging error variable. The network bias weight optimization vector is a key parameter set that indicates how to adjust the weights of each node in the decision tree to reduce the prediction error. The system uses a complex algorithm to convert the first network debugging error variable into this optimization vector, ensuring that each weight adjustment is based on actual error feedback.
[0162] If the error variable of the first network debugging is large, it means that the decision made by the deep semantic decision tree under the current weight configuration is significantly different from the actual result. At this time, the network bias weight optimization vector will be adjusted accordingly to guide the system on how to correct these deviations. On the contrary, if the error variable of the first network debugging is small, it means that the performance of the decision tree is relatively good, but the system will still make fine adjustments based on the suggestions of the optimization vector to pursue higher decision accuracy.
[0163] Next, the digital enterprise management system can optimize the network bias weight of the deep semantic decision tree based on the network bias weight optimization vector. This is a precise adjustment process. The system can adjust the weight of the corresponding node in the decision tree one by one according to each element in the optimization vector. In this way, the system can ensure that the decision tree is more accurate in future decisions and more in line with the actual business operations.
[0164] This optimization process involves not only numerical adjustments to weights, but may also include fine-tuning the decision tree structure. For example, the system may merge or split certain nodes based on the recommendations of the optimization vector to better adapt to the distribution and characteristics of the data. Such adjustments can further improve the generalization ability of the decision tree, enabling it to make accurate predictions and decisions when faced with new data.
[0165] In this way, the weight and structure of the deep semantic decision tree can be finely adjusted and optimized through the first network debugging error variable, which not only improves the decision accuracy of the decision tree, but also enhances its generalization ability and adaptability to new data. As the system continues to learn and optimize, it will be able to provide enterprises with more accurate and efficient decision support, thereby helping enterprises maintain their competitive advantage in a complex and changing market environment.
[0166] In some possible embodiments, the method involves a formula description of two key steps:
[0167] (1) According to the first network debugging error variable, the network bias weight optimization vector corresponding to the deep semantic decision tree is obtained. Assuming the first network debugging error variable is (E) and the network bias weight optimization vector is (q_W), this step can be expressed by the following formula:
[0168] [q_W=f(E)]
[0169] Among them, (f) is a function that calculates the weight optimization vector based on the network debugging error variable. This function can also involve complex mathematical operations and machine learning algorithms to ensure that the optimal weight adjustment solution is obtained based on the error variable.
[0170] (2) The network bias weight of the deep semantic decision tree is optimized according to the network bias weight optimization vector. Assuming that the original network bias weight of the deep semantic decision tree is (W), and the optimized network bias weight is (W'), the optimization process can be expressed by the following formula:
[0171] [W'=W+q_W]
[0172] This formula describes how to apply the calculated network bias weight optimization vector to the original network bias weight to obtain the optimized weight. This process is dynamically adjusted, and each optimization will update the weight according to the current error variable to improve the decision accuracy of the deep semantic decision tree.
[0173] In summary, these two formulas together constitute the mathematical model of the embodiment of the present application, which is used to dynamically optimize the network bias weights of the deep semantic decision tree according to the real-time network debugging error, thereby improving the decision performance.
[0174] In some other possible embodiments, the method also includes: obtaining a basic digital enterprise log analysis network; debugging the basic digital enterprise log analysis network based on the debugging digital enterprise log to obtain a current digital enterprise log analysis network; entering past digital enterprise logs into the current digital enterprise log analysis network for processing to obtain intermediate digital enterprise logs; when it is determined that the target abnormality warning level of the intermediate digital enterprise log relative to the past digital enterprise log is greater than the warning level threshold, jumping to the step of obtaining the first digital enterprise log in the first management task attention status.
[0175] In the above embodiments, the construction and optimization process of the log analysis network and how to use this network to detect and process abnormal logs are further considered.
[0176] First, the digital enterprise management system can obtain a basic digital enterprise log analysis network. This network is pre-built for preliminary analysis and processing of enterprise log data. It may contain a series of data processing modules, algorithms and models to extract key information from logs and identify potential problems or anomalies.
[0177] Next, the system can debug this basic network based on debugging digital enterprise logs. The purpose of debugging is to optimize the performance and accuracy of the network to ensure that it can better adapt to and process actual log data. The debugging process may include adjusting network parameters, optimizing model structure, updating algorithms, etc. After debugging, the system can obtain a current digital enterprise log analysis network, which has improved performance and accuracy.
[0178] The system can then enter past digital enterprise logs into this current analysis network for processing. These past logs may be data from a period of time before, containing various events and records in the enterprise's operations. Through the processing of the current analysis network, the system can obtain intermediate digital enterprise logs, which have been initially analyzed and screened, highlighting key information and potential problems.
[0179] In the process of processing the intermediate digital enterprise logs, the system can detect and determine the target abnormal warning level of these logs relative to the previous logs. If the abnormal warning level of a log exceeds the preset warning level threshold, it means that the log may contain important abnormal information or potential problems. At this time, the system can jump back to the step of obtaining the first digital enterprise log of the first management task focus state, so as to conduct a more in-depth analysis and processing of this abnormality.
[0180] In this way, an efficient and accurate log analysis network can be built to automatically detect and process abnormal information in enterprise logs, which not only improves the efficiency and accuracy of log analysis, but also helps enterprises to promptly discover and solve potential problems in the operation process, thereby ensuring the stable operation and sustainable development of the enterprise. At the same time, by continuously optimizing and debugging the log analysis network, the system can better adapt to and process the ever-changing log data, further improving its practicality and reliability in practical applications.
[0181] Under some other possible design ideas, determining the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record includes: obtaining a first thermal feature from the first semantic description thermal record, and obtaining a second thermal feature corresponding to the first thermal feature from the second semantic description thermal record; obtaining a characteristic distance between the first thermal feature and the second thermal feature; determining the semantic thermal correlation based on the characteristic distance, and the semantic thermal correlation has a first quantitative relationship with the characteristic distance.
[0182] In some design ideas, determining the semantic thermal correlation between two semantically described thermal records is a key step. This process aims to measure the similarity or correlation between two records for subsequent data analysis and processing.
[0183] First, the digital enterprise management system can extract the first thermal features from the first semantic description thermal record. These features may be keywords, phrases or data points closely related to the content of the record, which together constitute the unique identifier of the record. Similarly, the system will also extract the second thermal features corresponding to the first thermal features from the second semantic description thermal record.
[0184] Next, the system can calculate the feature distance between the two thermal features. This distance can be seen as the degree of "closeness" between the two features in the semantic space, which reflects the similarity or difference between the two records. The calculation of feature distance may involve complex mathematical formulas and algorithms, such as cosine similarity, Euclidean distance, etc., which are used to quantify the similarity between two features.
[0185] After obtaining the feature distance, the digital enterprise management system can determine the semantic thermal correlation between two semantically described thermal records based on this distance. This process is based on a preset first quantitative relationship, which may be a mathematical model or an empirical formula that maps the feature distance to the semantic thermal correlation. Simply put, the closer the distance between two features, the higher the semantic thermal correlation between them; conversely, the farther the distance, the lower the correlation.
[0186] In this way, the digital enterprise management system can accurately measure and compare the correlation between thermal records with different semantic descriptions, which is crucial for subsequent tasks such as data mining, information retrieval, and decision support. In this way, the accurate measurement of the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record can be achieved. This not only improves the accuracy and efficiency of the system in processing and analyzing large amounts of semantic data, but also provides enterprises with more accurate data insights and decision support. Specifically, this correlation analysis helps companies discover potential connections and trends hidden in the data, so as to make more informed business decisions. At the same time, this method can also be applied to multiple fields such as information retrieval and recommendation systems to help companies better understand and utilize their own data resources.
[0187] Under some preferred design ideas, the first log text memory semantic knowledge is obtained based on the first NLP semantic embedding algorithm, and the second log text memory semantic knowledge is obtained based on the second NLP semantic embedding algorithm. The network bias weight of the digital enterprise log analysis network is optimized according to the target network debugging error variable to obtain the debugged digital enterprise log analysis network, so as to use the debugged digital enterprise log analysis network to process the digital enterprise log, including: optimizing the network bias weights of the digital enterprise log analysis network, the first NLP semantic embedding algorithm and the second NLP semantic embedding algorithm according to the target network debugging error variable to obtain the optimized digital enterprise log analysis network, the first NLP semantic embedding algorithm and the second NLP semantic embedding algorithm; jumping to the step of obtaining the first digital enterprise log in the attention state of the first management task to continue network debugging until the digital enterprise log analysis network is in a stable state to obtain the debugged digital enterprise log analysis network.
[0188] In some preferred design ideas, there are more refined processing methods for the acquisition of log text memory semantic knowledge and the debugging process of digital enterprise log analysis networks.
[0189] First, the system can use two different NLP (natural language processing) semantic embedding algorithms to process log text data. Specifically, the first log text memory semantic knowledge is obtained through the first NLP semantic embedding algorithm, which can convert the log text into a representation in a high-dimensional vector space to capture the semantic information of the text. Similarly, the second log text memory semantic knowledge is obtained based on the second NLP semantic embedding algorithm. These two algorithms may have different characteristics and advantages in order to more comprehensively extract the semantic features of the log text.
[0190] Next, in the process of optimizing the digital enterprise log analysis network, the system not only considers the bias weight of the network itself, but also optimizes the network bias weights related to the first NLP semantic embedding algorithm and the second NLP semantic embedding algorithm. This comprehensive optimization method can ensure that the log analysis network can achieve higher accuracy and efficiency when processing log text embedded by the NLP algorithm.
[0191] The optimization process is based on the target network debugging error variable. The system can analyze the errors in the network debugging process and adjust the network bias weights of each algorithm accordingly to reduce the errors and improve the performance of the analyzed network. This process may involve complex mathematical calculations and algorithm iterations to ensure that the weight adjustment can maximize the overall performance of the network.
[0192] When the network bias weight is optimized, the system can jump back to the step of obtaining the first digital enterprise log of the first management task focus state to continue network debugging. This process is iterative, and the system can continuously optimize the network weight according to the debugging error until the digital enterprise log analysis network reaches a stable state. In this state, the network can accurately process and analyze digital enterprise logs, providing valuable insights and decision support for the enterprise.
[0193] In this way, digital enterprise logs can be processed and analyzed more accurately, providing more accurate and comprehensive semantic knowledge. This comprehensive optimization method not only improves the performance and accuracy of the log analysis network, but also enables the system to better adapt and utilize different NLP semantic embedding algorithms, thereby more comprehensively capturing the semantic information of log text. Ultimately, this will help enterprises use log data more efficiently, make more informed decisions, and promote digital transformation and upgrading of enterprises.
[0194] Based on the above embodiments / design ideas, the method includes: obtaining a pending digital enterprise log for which an attention status is to be updated; determining a first management task attention status matched by the pending digital enterprise log and a second management task attention status to be updated; determining a debugged digital enterprise log analysis network based on the first management task attention status and the second management task attention status; the digital enterprise log analysis network is debugged based on a target network debugging error variable, the target network debugging error variable is obtained based on a first network debugging error variable and a second network debugging error variable, the first network debugging error variable is determined based on a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and the first network debugging error variable and the target memory semantic commonality score have a first quantitative relationship, the second network debugging error variable is obtained according to the semantic thermal correlation between the first semantic description thermal record corresponding to the first log text memory semantic knowledge and the second semantic description thermal record corresponding to the second log text memory semantic knowledge, the second network debugging error variable and the semantic thermal correlation have a first quantitative relationship, the first log text memory semantic knowledge is the log text memory semantic knowledge corresponding to the first digital enterprise log of the first management task attention state, the second log text memory semantic knowledge is the log text memory semantic knowledge corresponding to the second digital enterprise log of the second management task attention state, and the second digital enterprise log is obtained based on the first digital enterprise log; the digital enterprise log to be processed is entered into the debugged digital enterprise log analysis network, and the target digital enterprise log of the second management task attention state is output.
[0195] In a digital enterprise management system, in order to continuously update and optimize the log analysis network and better adapt to the changing management task requirements, the system can perform a series of complex operations.
[0196] First, the system can obtain pending digital enterprise logs that are to be updated. These logs may come from various departments and business processes of the enterprise and contain a large amount of information related to the enterprise's operations.
[0197] Next, the system can determine the first management task attention status that matches these pending logs and the second management task attention status that needs to be updated. The management task attention status reflects the importance and processing priority of different log information by the enterprise. For example, some logs may be related to key business processes and require higher attention.
[0198] In order to accurately process these logs, the system can determine the debugged digital enterprise log analysis network according to the first and second management task attention states. This network is obtained through a series of debugging processes to ensure that it can effectively process and analyze various log data.
[0199] During the debugging process, the system can consider multiple factors to determine the target network debugging error variable. This includes the first network debugging error variable determined based on the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and the second network debugging error variable determined based on the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record. These scores and correlations reflect the similarity and correlation between different log texts, which are crucial for optimizing the log analysis network.
[0200] Specifically, the first log text memory semantic knowledge and the second log text memory semantic knowledge correspond to the first digital enterprise log of the first management task focus state and the second digital enterprise log of the second management task focus state respectively. The system can extract and calculate the memory semantic commonality score and semantic thermal correlation based on the semantic features of these log texts, thereby guiding the debugging and optimization of the network.
[0201] Finally, when the debugged digital enterprise log analysis network is ready, the system can enter the digital enterprise logs to be processed into the network for analysis and processing. Through the complex calculation and processing flow of the network, the system can output the target digital enterprise logs corresponding to the state of concern of the second management task. These logs have been deeply analyzed and processed, and can directly provide strong support for the management decision-making of the enterprise.
[0202] Through the above methods, the digital enterprise management system can achieve accurate processing and analysis of log data. By continuously optimizing and debugging the log analysis network, the system can more accurately identify and extract key information, and improve the efficiency and accuracy of log processing. At the same time, the network parameters and analysis strategies are adjusted according to the attention status of different management tasks, so that the system can better meet the changing needs of enterprises.
[0203] In some independent embodiments, after entering the to-be-processed digital enterprise log into the debugged digital enterprise log analysis network and outputting the target digital enterprise log of the second management task attention state, the method further includes:
[0204] Obtain abnormal risk view annotation logs;
[0205] Obtaining the risk event analysis results in the target digital enterprise log and the risk event analysis results in the abnormal risk viewpoint annotation log;
[0206] Determine the risk event commonality weight of the target digitized enterprise log and the abnormal risk viewpoint annotated log according to the risk event analysis results in the target digitized enterprise log and the risk event analysis results in the abnormal risk viewpoint annotated log;
[0207] Determining structured text features of the target digitized enterprise log and structured text features of the abnormal risk viewpoint annotated log;
[0208] Determining a first structured text commonality weight of the target digitized enterprise log and the abnormal risk viewpoint annotated log according to the structured text features of the target digitized enterprise log and the structured text features of the abnormal risk viewpoint annotated log;
[0209] Determining the semantic relevance between the target digitized enterprise log and the abnormal risk viewpoint annotated log according to the risk event commonality weight of the target digitized enterprise log and the abnormal risk viewpoint annotated log and the first structured text commonality weight;
[0210] According to the semantic relevance between the target digitized enterprise log and the abnormal risk viewpoint annotation log, the target digitized enterprise log is text-annotated using the abnormal risk viewpoint annotation log.
[0211] In some independent embodiments of the digital enterprise management system, the system not only processes and analyzes the logs, but also further annotates the logs with abnormal risk perspectives to enhance the information content and accuracy of the target digital enterprise logs.
[0212] First, the system can obtain abnormal risk opinion annotation logs. These logs usually contain annotations and opinions on specific risk events, which are recorded by experts or experienced operators based on actual conditions.
[0213] Next, the system can obtain the risk event analysis results in the target digital enterprise log and the abnormal risk view annotation log respectively. The risk event analysis result is a detailed analysis and interpretation of the risk events mentioned in the log, which includes information such as the type of risk, severity, and possible impact range.
[0214] Then, the system can determine the risk event commonality weight between the two logs based on the risk event analysis results in the two logs. This weight reflects the similarity and overlap in the risk event descriptions of the two logs. The higher the weight, the more consensus the two logs have on the risk events.
[0215] In addition to the risk event commonality weights, the system can also determine the structured text features of the two logs and calculate the first structured text commonality weight between them based on these features. The structured text features may include the writing format of the logs, the vocabulary used, the sentence structure, etc., and the first structured text commonality weight quantifies the similarity of these features in the two logs.
[0216] With the risk event commonality weight and the first structured text commonality weight, the system can combine these two weights to determine the semantic relevance between the target digital enterprise log and the abnormal risk viewpoint annotated log. This semantic relevance is a comprehensive measure that considers the similarity of the logs in terms of risk events and structured text features.
[0217] Finally, the system can use the abnormal risk viewpoint annotation log to perform text annotation on the target digital enterprise log based on this semantic relevance. The annotation process may be to add relevant information (such as risk type, solution, etc.) in the abnormal risk viewpoint annotation log to the target digital enterprise log, or to label the target digital enterprise log with a label similar to the abnormal risk viewpoint annotation log. In this way, the target digital enterprise log obtains richer information and more accurate annotations, which helps enterprises better identify and manage risks.
[0218] Through the above embodiments, the digital enterprise management system can make full use of the information of the abnormal risk viewpoint annotated log to enhance the accuracy and information content of the target digital enterprise log. This cross-log analysis and annotation method not only improves the utilization value of the log, but also helps enterprises to understand the risk situation more comprehensively, so as to make more timely and accurate decisions. In addition, by comprehensively considering the similarities between risk events and structured text features, the system can more accurately identify the correlation between different logs, further improving the depth and breadth of log analysis.
[0219] Further, Figure 2 This is a schematic diagram of the structure of a digital enterprise management system 200 provided in an embodiment of the present application. Figure 2 The digital enterprise management system 200 shown includes a processor 210, which can call and run a computer program from a memory to implement the method in the embodiment of the present application.
[0220] Alternatively, if Figure 2 As shown, the digital enterprise management system 200 may further include a memory 230. The processor 210 may call and run a computer program from the memory 230 to implement the method in the embodiment of the present application.
[0221] The memory 230 may be a separate device independent of the processor 210 , or may be integrated into the processor 210 .
[0222] Alternatively, if Figure 2 As shown, the digital enterprise management system 200 may further include a transceiver 220, and the processor 210 may control the transceiver 220 to interact with other devices, specifically, to send information or data to other devices, or to receive information or data sent by other devices.
[0223] Optionally, the digital enterprise management system 200 may implement the corresponding processes corresponding to the storage engine or components in the storage engine (such as a processing module) or a device deployed with a storage engine in each method of the embodiments of the present application, which will not be described in detail here for the sake of brevity.
[0224] It should be understood that the processor of the embodiment of the present application may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiment can be completed by the hardware integrated logic circuit or software instructions in the processor. The above processor can be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined and performed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.
[0225] It can be understood that the memory in the embodiments of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DR RAM). It should be noted that the memory of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.
[0226] It should be understood that the above-mentioned memory is exemplary but not restrictive. For example, the memory in the embodiments of the present application may also be static random access memory (static RAM, SRAM), dynamic random access memory (dynamic RAM, DRAM), synchronous dynamic random access memory (synchronous DRAM, SDRAM), double data rate synchronous dynamic random access memory (double data rate SDRAM, DDR SDRAM), enhanced synchronous dynamic random access memory (enhanced SDRAM, ESDRAM), synchronous link dynamic random access memory (synch link DRAM, SLDRAM) and direct memory bus random access memory (Direct Rambus RAM, DR RAM), etc. That is to say, the memory in the embodiments of the present application is intended to include but not limited to these and any other suitable types of memory.
[0227] Based on the above, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program implements the above method when running.
[0228] 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 modifications and variations.
Claims
1. A digital enterprise management method based on machine learning, characterized in that: Applied to a digital enterprise management system, the method comprises: Acquire a first digital enterprise log of a first management task attention state, enter the first digital enterprise log into a digital enterprise log analysis network to be debugged, and obtain a second digital enterprise log of a second management task attention state; Performing text memory semantic recognition on the first digitized enterprise log to obtain first log text memory semantic knowledge; performing text memory semantic recognition on the second digitized enterprise log to obtain second log text memory semantic knowledge; Determine a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, obtain a first network debugging error variable according to the target memory semantic commonality score, and the first network debugging error variable has a first quantitative relationship with the target memory semantic commonality score; Acquire a first semantic description thermal record corresponding to the first log text memory semantic knowledge and a second semantic description thermal record corresponding to the second log text memory semantic knowledge; Determine a semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record, obtain a second network debugging error variable according to the semantic thermal correlation, and the second network debugging error variable has a first quantitative relationship with the semantic thermal correlation; Obtaining a target network debugging error variable according to the first network debugging error variable and the second network debugging error variable; Optimizing the network bias weight of the digital enterprise log analysis network according to the target network debugging error variable to obtain a debugged digital enterprise log analysis network, so as to process the digital enterprise logs using the debugged digital enterprise log analysis network; The target memory semantic commonality score includes a first semantic decision possibility, and the determining of the target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, and obtaining a first network debugging error variable according to the target memory semantic commonality score includes: combining the first log text memory semantic knowledge and the second log text memory semantic knowledge, and using the combined semantic knowledge as a positive example of memory semantic knowledge; Entering the positive examples of the memorized semantic knowledge into a deep semantic decision tree to obtain a first semantic decision possibility; Obtaining a first network debugging error variable according to the first semantic decision possibility; Wherein, obtaining the first network debugging error variable according to the first semantic decision possibility includes: obtaining a third digitized enterprise log of the attention status of the second management task, the third digitized enterprise log is different from the second digitized enterprise log; performing text memory semantic recognition on the third digitized enterprise log to obtain third log text memory semantic knowledge; taking the target log text memory semantic knowledge and the third log text memory semantic knowledge as negative examples of memory semantic knowledge, entering the negative examples of memory semantic knowledge into a deep semantic decision tree to obtain a second semantic decision possibility, the target log text memory semantic knowledge is the first log text memory semantic knowledge or the second log text memory semantic knowledge; obtaining the first network debugging error variable according to the first semantic decision possibility and the second semantic decision possibility, the second semantic decision possibility and the first network debugging error variable are in a second quantitative relationship; Among them, the method also includes: obtaining a network bias weight optimization vector corresponding to the deep semantic decision tree based on the first network debugging error variable; and optimizing the network bias weight of the deep semantic decision tree based on the network bias weight optimization vector.
2. The method according to claim 1, characterized in that The step of obtaining the first semantic description thermal record corresponding to the first log text memory semantic knowledge comprises: Performing a thermal analysis based on a first semantic description value in the first log text memory semantic knowledge to obtain a first thermal distribution coefficient corresponding to the first semantic description value; A first semantic description thermal record corresponding to the first log text memory semantic knowledge is determined based on the first thermal distribution coefficient.
3. The method according to claim 2, characterized in that The first thermal distribution coefficient includes a target global indicator and a target discrete indicator, and determining a first semantic description thermal record corresponding to the first log text memory semantic knowledge based on the first thermal distribution coefficient includes: Acquire a target indication weight set, wherein the target indication weight set includes a plurality of log semantic reconstruction weights; Performing an adjustment operation on the target discrete indicator based on the log semantic reconstruction weight to obtain a target semantic adjustment indicator; The target global indicator is modified according to the target semantic adjustment indicator to obtain a target log event label value, and the target log event label values corresponding to each log semantic reconstruction weight constitute a target log event label value set; Determine the target activity index corresponding to each target log event tag value based on the target log event tag value set, obtain a first semantic description thermal record based on the target activity index corresponding to each target log event tag value, and use the target activity index corresponding to each target log event tag value as the thermal feature in the first semantic description thermal record.
4. The method according to claim 3, characterized in that The target activity index is determined according to the global semantic thermal spectrum corresponding to the target log event tag value set; The acquiring target indication weight set includes a plurality of log semantic reconstruction weights including: Acquire semantic thermal reference conditions, extract log event label values according to the semantic thermal reference conditions, and obtain log event label values corresponding to the semantic thermal reference conditions as log semantic reconstruction weights.
5. The method according to claim 1, characterized in that The method further comprises: Get the basic digital enterprise log analysis network; Debugging the basic digital enterprise log analysis network based on the debugging digital enterprise log to obtain the current digital enterprise log analysis network; Entering past digital enterprise logs into the current digital enterprise log analysis network for processing to obtain intermediate digital enterprise logs; When it is determined that the target abnormality warning level of the intermediate digitized enterprise log relative to the past digitized enterprise log is greater than the warning level threshold, the process jumps to the step of obtaining the first digitized enterprise log in the first management task attention state.
6. The method according to claim 1, characterized in that The determining of the semantic thermal correlation between the first semantic description thermal record and the second semantic description thermal record comprises: Acquire a first thermal feature from the first semantic description thermal record, and acquire a second thermal feature corresponding to the first thermal feature from the second semantic description thermal record; Acquire a characteristic distance between the first thermal characteristic and the second thermal characteristic; A semantic thermal correlation is determined based on the feature distance, wherein the semantic thermal correlation has a first quantitative relationship with the feature distance.
7. The method according to claim 1, characterized in that The first log text memory semantic knowledge is obtained based on a first NLP semantic embedding algorithm, the second log text memory semantic knowledge is obtained based on a second NLP semantic embedding algorithm, and the step of optimizing the network bias weight of the digital enterprise log analysis network according to the target network debugging error variable to obtain a debugged digital enterprise log analysis network, so as to process the digital enterprise log using the debugged digital enterprise log analysis network includes: Optimizing the network bias weights of the digital enterprise log analysis network, the first NLP semantic embedding algorithm, and the second NLP semantic embedding algorithm according to the target network debugging error variable to obtain optimized digital enterprise log analysis network, the first NLP semantic embedding algorithm, and the second NLP semantic embedding algorithm; Jump to the step of obtaining the first digital enterprise log of the first management task attention state to continue network debugging until the digital enterprise log analysis network is in a stable state, and obtain the debugged digital enterprise log analysis network.
8. The method according to claim 1, characterized in that The method further comprises: Obtain pending digital enterprise logs for which attention status updates are to be made; Determine a first management task attention status matched by the to-be-processed digital enterprise log and a second management task attention status to be updated; Determine a debugged digital enterprise log analysis network according to the first management task attention state and the second management task attention state; the digital enterprise log analysis network is debugged based on a target network debugging error variable, the target network debugging error variable is obtained according to the first network debugging error variable and the second network debugging error variable, the first network debugging error variable is determined according to a target memory semantic commonality score between the first log text memory semantic knowledge and the second log text memory semantic knowledge, the first network debugging error variable and the target memory semantic commonality score have a first quantitative relationship, the second network debugging error variable is obtained according to a semantic thermal correlation between a first semantic description thermal record corresponding to the first log text memory semantic knowledge and a second semantic description thermal record corresponding to the second log text memory semantic knowledge, the second network debugging error variable and the semantic thermal correlation have a first quantitative relationship, the first log text memory semantic knowledge is a log text memory semantic knowledge corresponding to a first digital enterprise log of the first management task attention state, the second log text memory semantic knowledge is a log text memory semantic knowledge corresponding to a second digital enterprise log of the second management task attention state, and the second digital enterprise log is obtained according to the first digital enterprise log; The digital enterprise log to be processed is entered into the debugged digital enterprise log analysis network, and the target digital enterprise log of the second management task focus status is output.
9. A digital enterprise management system, characterized in that: The method comprises at least one processor and a memory; the memory stores computer-executable instructions; the at least one processor executes the computer-executable instructions stored in the memory, so that the at least one processor executes the method according to any one of claims 1 to 8.
