Artificial intelligence-based document conversion service stability optimization method and system

Abstract

The application provides a document conversion service stability optimization method and system based on artificial intelligence, acquires a real-time running state data set containing service process running data, resource occupation data and conversion task processing data, obtains a service running state feature set through feature extraction, then calls a pre-trained service anomaly tracing module to identify an abnormal mode, generates a result containing an abnormal type identifier and abnormal correlation features, calls a corresponding service adaptation model based on the result and adjusts parameters, generates a service stability regulation strategy, inputs the strategy into an execution control module for execution, obtains a service running parameter adjustment result, and finally updates a training sample set of the service adaptation model according to the adjustment result, iteratively optimizes to generate an optimized service adaptation model. The application realizes comprehensive optimization of the stability of the document conversion service through multi-link cooperation, and improves the service performance and reliability.

Description

Technical Field

[0001] This invention relates to the field of digital office and information processing, and more specifically, to a method and system for optimizing the stability of document conversion services based on artificial intelligence. Background Technology

[0002] In the field of digital office and information processing, document conversion services play a crucial role, enabling the conversion between documents of different formats to meet diverse business needs. However, as the application scenarios of document conversion services continue to expand and the processing tasks become increasingly complex, their stability faces numerous challenges.

[0003] Existing methods for ensuring the stability of document conversion services have significant shortcomings. On the one hand, some methods rely heavily on manual experience for monitoring and adjustment. When service anomalies occur, maintenance personnel need to manually analyze large amounts of operational data, which is not only inefficient but also prone to overlooking critical issues due to limitations in personal experience and judgment, leading to untimely anomaly handling and impacting service stability. On the other hand, while some automated monitoring methods can collect service operation data in real time, they lack intelligent anomaly identification and control mechanisms. They can only issue simple threshold alarms and cannot accurately identify anomaly types and related characteristics, thus failing to formulate targeted stability control strategies and fundamentally solve service stability problems. Furthermore, existing methods do not provide effective feedback and optimization of the anomaly handling process after handling service anomalies, leading to the recurrence of similar problems, further reducing service stability and reliability. A comprehensive optimization approach, considering factors such as format and attributes, is needed to improve the overall effect, thereby affecting the quality and efficiency of electronic document printing. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an artificial intelligence-based document conversion service stability optimization method, the method comprising:

[0005] Obtain a real-time running status data set of the document conversion service, perform feature extraction processing on the real-time running status data set to obtain a service running status feature set, wherein the real-time running status data set includes service process running data, resource usage data and conversion task processing data;

[0006] The pre-trained service anomaly tracing module is invoked to perform anomaly pattern recognition processing on the service running status feature set, generating service anomaly pattern recognition results, which include anomaly type identifiers and anomaly association features.

[0007] Based on the service anomaly pattern recognition result, the corresponding service adaptation model is invoked, the parameters of the service adaptation model are adjusted, and a service stability control strategy is generated. The service stability control strategy includes a resource allocation scheme and task scheduling rules.

[0008] The service stability control strategy is input into the execution control module of the document conversion service, and the service stability control strategy is executed to obtain the service operation parameter adjustment results;

[0009] The training sample set of the service adaptation model is updated based on the adjustment results of the service operation parameters, and the service adaptation model is iteratively optimized to generate an optimized service adaptation model.

[0010] In another aspect, embodiments of the present invention also provide an artificial intelligence-based document conversion service stability optimization system, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0011] Based on the above, this embodiment of the invention extracts features by merging the real-time running status dataset of the document conversion service and calling a pre-trained service anomaly tracing module for anomaly pattern recognition. This accurately identifies anomaly types and associated features, avoiding the subjectivity and limitations of manual judgment and improving the accuracy and efficiency of anomaly identification. Based on the identification results, the corresponding service adaptation model is invoked to generate a service stability control strategy. This allows for the development of targeted resource allocation schemes and task scheduling rules according to different anomaly types, achieving precise control and effectively improving service stability. The control strategy is input into the execution control module to obtain the results of the running parameter adjustments. Finally, based on the results of the running parameter adjustments, the training sample set of the service adaptation model is updated and iteratively optimized, enabling the model to continuously learn and adapt to changes in service operation, further improving the accuracy of anomaly identification and control, and significantly enhancing the stability and reliability of the document conversion service. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the execution flow of the document conversion service stability optimization method based on artificial intelligence provided in an embodiment of the present invention.

[0013] Figure 2 This is a schematic diagram of exemplary hardware and software components of the document conversion service stability optimization system based on artificial intelligence provided in this embodiment of the invention. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1This is a flowchart illustrating an AI-based document conversion service stability optimization method according to an embodiment of the present invention. The following is a detailed description of this AI-based document conversion service stability optimization method.

[0015] Step S110: Obtain the real-time running status data set of the document conversion service, perform feature extraction processing on the real-time running status data set to obtain the service running status feature set, wherein the real-time running status data set includes service process running data, resource usage data and conversion task processing data.

[0016] In this embodiment, an enterprise internal document conversion service system is used as the application scenario. It is deployed on the enterprise's private cloud server to handle various document format conversion needs generated by employees in their daily work. It supports mutual conversion between multiple document types, such as text documents, table documents, presentation documents, and other different types of document format conversion operations.

[0017] Step S111: Collect service process running data generated during the operation of the document conversion service. The service process running data includes process startup time data, process response delay data, and process interaction data.

[0018] This enterprise document conversion service system runs multiple independent service processes, each corresponding to a specific type of document conversion function, such as a process specifically for text document conversion, a process specifically for table document conversion, etc. Process startup time data refers to the time interval from the system startup command being issued to the process being fully ready to receive conversion tasks. This data indicates the process startup efficiency. Process response latency data refers to the time interval from the service process receiving a document conversion request from the client to starting the conversion operation. This response latency data directly relates to the user's perception of service response speed. Process interaction data covers communication information between the service process and other components within the system (such as the task scheduling module and resource management module), including the initiation time of the interaction request, the type of interaction data, and the result status of the interaction operation.

[0019] Step S112: Collect resource usage data during the operation of the document conversion service, including memory usage data, processor usage data, and network bandwidth usage data.

[0020] This enterprise document conversion service system operates in a cluster environment composed of multiple servers. Resource usage data is collected for each server and the service processes running within it. Memory usage data includes the amount of memory currently allocated to each service process and the actual amount of memory used. Different types of document conversion tasks have different memory requirements. For example, processing presentation documents containing many images typically consumes more memory than processing plain text documents. Processor usage data reflects the utilization of server CPU resources by the service processes, including the processor usage ratio and the processor time slice allocation between different service processes. When multiple highly complex conversion tasks are executed concurrently, processor utilization will increase significantly. Network bandwidth usage data involves the network resources used by the service processes in receiving document data to be converted from the client and returning the converted document data to the client. This includes indicators such as data transmission rate, number of network connections, and packet loss rate during data transmission. In an enterprise internal network environment, the allocation of network bandwidth resources affects the overall processing efficiency of the document conversion service.

[0021] Step S113: Collect conversion task processing data when the document conversion service processes tasks. The conversion task processing data includes task reception quantity data, task processing progress data, and task completion time data.

[0022] Internal employees submit various conversion tasks to the document conversion service system via a client application. Task reception data records the total number of conversion tasks received by the service system per unit of time, as well as the proportion of different document conversion task types. The trend of this task reception data reflects the service load. Task processing progress data refers to the real-time progress of each conversion task during execution, such as the percentage of document pages that have been converted out of the total pages, and the currently processed document paragraphs. This data allows for real-time monitoring of task execution status. Task completion time data refers to the total time consumed from when the conversion task is received by the service process and begins processing until the task is fully completed. This task completion time data is related to factors such as document size, complexity, and format type. Analyzing this data allows for the evaluation of the processing efficiency of different types of conversion tasks.

[0023] Step S114: Integrate the service process running data, the resource usage data, and the conversion task processing data according to the time series to form a real-time running status data set. Perform feature extraction processing on the real-time running status data set. Extract process running features, resource usage features, and task processing features through feature extraction algorithms. The process running features correspond to the key attributes of the service process running data, the resource usage features correspond to the key attributes of the resource usage data, and the task processing features correspond to the key attributes of the conversion task processing data.

[0024] During the data integration phase, the collected service process operation data, resource usage data, and conversion task processing data are associated and combined according to timestamps to form a real-time operation status data set with time as the axis. For example, process response latency data at a specific time point, memory usage data at that time point, and task completion time data of the conversion task being processed at that time are associated and stored. The feature extraction process employs a combination of multiple feature extraction algorithms. For process operation data, by analyzing the changing patterns of process startup time data, the distribution characteristics of process response latency data, and key events in process interaction data, process operation features reflecting process stability, response timeliness, and collaborative work capabilities are extracted. For resource usage data, by analyzing the fluctuations in memory usage data, the peak and trough changes in processor usage data, and the transmission efficiency of network bandwidth usage data, resource usage features reflecting resource utilization efficiency and rational resource allocation are extracted. For conversion task processing data, by analyzing the time distribution characteristics of task reception quantity data, the progress rate of task processing data, and the statistical patterns of task completion time data, task processing features reflecting task load pressure and task processing efficiency are extracted.

[0025] Step S115: After the process running features, resource occupancy features and task processing features are input into the feature fusion module for standardization, the feature splicing and weight allocation are combined for fusion processing to generate a service running status feature set with a unified dimension.

[0026] First, the extracted process execution features, resource consumption features, and task processing features are input into the feature fusion module for standardization. During standardization, for different types of features, based on their data distribution range and feature attributes, the feature values ​​are transformed to a unified numerical range to eliminate the influence of differences in units between different features. For example, process response latency and memory usage features are standardized separately to ensure their numerical ranges are within the same interval. After standardization, feature fusion is performed. Feature concatenation combines process execution features, resource consumption features, and task processing features into a high-dimensional feature vector according to a preset order; weight allocation assigns corresponding weight values ​​to each feature based on its importance in reflecting the stability of service operation status. The assessment of importance is based on the analysis results of the contribution of each feature to the occurrence of anomalies in historical service anomaly cases. During fusion, each feature is first multiplied by its corresponding weight value, and then all the weighted features are combined according to the concatenation order to finally generate a set of service operation status features with a unified dimension.

[0027] Step S120: Call the pre-trained service anomaly tracing module to perform anomaly pattern recognition processing on the service running status feature set, and generate service anomaly pattern recognition results. The service anomaly pattern recognition results include anomaly type identifiers and anomaly association features.

[0028] After obtaining the service runtime status feature set, it is necessary to perform anomaly pattern recognition to determine whether the service currently has anomalies, as well as the specific types and related characteristics of the anomalies. The pre-trained service anomaly tracing module is trained based on a large amount of historical service runtime data and anomaly cases, and has the ability to perform in-depth analysis and pattern recognition of service runtime status features. By calling this service anomaly tracing module to process the input service runtime status feature set, it is possible to accurately identify the anomaly patterns that may exist during service operation.

[0029] Step S121: Input the service operation status feature set into the feature layering unit of the service anomaly tracing module, and perform layering processing on the service operation status feature set according to the data source and correlation to obtain the basic feature layer and the related feature layer. The basic feature layer contains features corresponding to a single type of data, namely, independent features corresponding to service process operation data, resource consumption data, and conversion task processing data respectively. The related feature layer contains the correlation features between different types of data, namely, the correlation features between service process operation data and resource consumption data, the correlation features between resource consumption data and conversion task processing data, and the correlation features between service process operation data and conversion task processing data.

[0030] The feature layering unit of the service anomaly tracing module first parses the input service runtime status feature set and divides it into different basic feature sub-layers based on the data source of the features. For example, features related to service process operation (such as process response latency features, process startup time features, etc.) are divided into the process basic feature sub-layer; features related to resource consumption (such as memory consumption features, processor consumption features, etc.) are divided into the resource basic feature sub-layer; and features related to task processing (such as task completion time features, number of tasks received features, etc.) are divided into the task basic feature sub-layer. These sub-layers together constitute the basic feature layer. The construction of the associated feature layer is achieved by analyzing the inherent correlation between features in different basic feature sub-layers. For example, we can analyze the relationship between process response latency in the process basic feature sublayer and processor usage in the resource basic feature sublayer to see if process response latency increases when processor usage is high, thus constructing a correlation feature between the two. We can also analyze the relationship between memory usage in the resource basic feature sublayer and task completion time in the task basic feature sublayer to see if task completion time changes significantly when memory usage reaches a certain level, thus forming a corresponding correlation feature. Finally, we can analyze the relationship between process interaction data in the process basic feature sublayer and task processing progress in the task basic feature sublayer to see if the smoothness of process interaction affects the progress of task processing, thus constructing a correlation feature.

[0031] Step S122: Perform anomaly threshold comparison processing on each feature in the basic feature layer. Through the preset normal range of features, filter out features that exceed the normal range of features as basic abnormal features. Perform correlation degree analysis processing on each associated feature in the associated feature layer. By calculating the standardized correlation strength value of the associated features, filter out features whose standardized correlation strength value exceeds the normal correlation range as associated abnormal features.

[0032] For each feature in the basic feature layer, a normal range is preset based on its value range under historical normal operating conditions. During the anomaly threshold comparison process, the current feature value is compared with the preset normal range. If the feature value exceeds this range, it is marked as a basic anomalous feature. For example, for the process response delay feature, a normal range is preset; when the feature value exceeds the upper limit of this range at a certain moment, it is determined to be a basic anomalous feature. For each associated feature in the associated feature layer, the strength of the association is evaluated by calculating its standardized association strength value. The calculation of the standardized association strength value is based on the distribution of the association strength of the feature under normal operating conditions in historical data, converting the current association strength value into a standardized value. A normal association range is preset; when the standardized association strength value exceeds this range, it indicates that the association between the corresponding features is abnormal, and the associated feature is marked as an anomalous association feature. For example, for the association feature between processor usage and task completion time, when the calculated standardized association strength value exceeds the normal association range, it indicates that the relationship between processor usage and task completion time is abnormal, and the associated feature is filtered as an anomalous association feature.

[0033] Step S123: Input the basic anomaly features and the associated anomaly features into the pattern matching unit of the service anomaly tracing module, call the preset anomaly pattern library, compare the basic anomaly features with the standard basic anomaly features one by one, and compare the associated anomaly features with the standard associated anomaly features one by one. The anomaly pattern library stores a set of standard anomaly features corresponding to various historical service anomalies. Each set of standard anomaly features includes standard basic anomaly features and standard associated anomaly features.

[0034] Step S1231: Call the preset abnormal pattern library, read all standard abnormal feature sets stored in the abnormal pattern library, and identify each standard abnormal feature set.

[0035] In this embodiment, the exception pattern library adopts a distributed storage architecture, deployed on multiple nodes of the enterprise private cloud, and accessed through a unified interface. When invoking the preset exception pattern library, a query request is first sent through the interface, containing instructions to retrieve all standard exception feature sets. Upon receiving the request, the exception pattern library performs authentication and permission checks. After successful authentication, it reads the metadata information of all standard exception feature sets from the distributed storage nodes, including the set name, corresponding exception type description, creation time, and update time. Then, based on this metadata information, each standard exception feature set is uniquely identified. The identifier uses a combination of letters and numbers, where the letters represent the major category of the exception type, and the numbers represent the specific exception subclass and version information under that major category. For example, an identifier starting with "MEM" represents an exception type related to memory resources, and the subsequent numbers represent the specific exception subclass and the updated version of the standard feature set of that subclass. After identification, the identifiers of all standard exception feature sets and their corresponding metadata information are returned to the pattern matching unit, which stores the above information in a local cache for subsequent fast access.

[0036] Step S1232: For each set of standard anomaly features, compare each feature in the set of standard anomaly features with the corresponding standard basic anomaly features in terms of feature attributes. Before comparison, standardize all feature values ​​and calculate the standardized difference value between the two in the attribute dimension. The standardized difference value is negatively correlated with the degree of matching. The degree of matching corresponding to the standardized difference value conforms to the preset association relationship. At the same time, calculate the similarity between the numerical distribution of the standardized basic anomaly features and the numerical distribution of the standardized standard basic anomaly features to obtain the basic feature similarity.

[0037] For each set of standard anomaly features, the pattern matching unit reads its corresponding list of standard basic anomaly features from the local cache. For each feature in the basic anomaly features list, the corresponding standard basic anomaly feature is searched based on the feature name and attribute description. After finding the corresponding feature, the feature values ​​of both are first standardized. The standardization process maps the feature value to a preset standard numerical range. The mapping method is determined based on the maximum and minimum values ​​of the feature under historical normal operating conditions. The standardized feature value is obtained by subtracting the historical minimum value from the current feature value and dividing by the difference between the historical maximum and minimum values. After standardization, feature attribute comparison is performed. Attribute comparison includes the data type, dimension (based on the previously unified dimension, this mainly checks the dimension consistency indicator), and physical meaning description of the feature. For numerical attributes, the standardized difference value between the two is calculated in the attribute dimension. The calculation method is to subtract the standardized value of the basic anomaly feature from the standardized value of the standard basic anomaly feature and take the absolute value. The smaller the standardized difference value, the higher the degree of matching between the two features in the numerical attribute; that is, the standardized difference value is negatively correlated with the degree of matching. The pre-defined association relationships are determined by looking up a pre-defined mapping table, which stores the matching degree levels corresponding to different standardized difference value intervals. For example, a standardized difference value corresponds to a matching degree level of "high" in one interval and "medium" in another. Simultaneously, the similarity between the numerical distribution of the standardized basic anomaly features and the numerical distribution of the standard basic anomaly features is calculated. The similarity calculation of the numerical distributions uses statistical methods. First, the probability density functions of the standardized values ​​of the two features are calculated separately. The probability density function is calculated based on the kernel density estimation method, by selecting an appropriate kernel function and bandwidth parameter to fit the standardized feature value samples. Then, the integral probability distance between the two probability density functions is calculated. The smaller the integral probability distance, the more similar the numerical distributions of the two features are. This integral probability distance is converted into a similarity index, which is the basic feature similarity. The basic feature similarity comprehensively considers the matching degree level of feature attribute comparison and the similarity index of numerical distributions, and is obtained through a weighted summation. The weights are determined according to the importance of the feature attributes; for example, the matching weight of features describing physical meaning is higher than that of numerical attributes.

[0038] Step S1233: For each set of standard anomaly features, compare the correlation relationship between each feature in the set of standard anomaly features and the corresponding standard correlation anomaly feature, analyze the consistency between the two in terms of correlation direction and correlation strength change trend, calculate the degree of agreement of the correlation relationship, and at the same time calculate the similarity of the time series change pattern of the correlation features to obtain the correlation feature similarity.

[0039] For each set of standard anomaly features, the pattern matching unit also identifies the corresponding associated anomaly features based on the feature name and association description. The association comparison first analyzes the direction of association, categorized into positive, negative, and no correlation. This is determined by comparing the association direction indicators of the basic anomaly features and the standard basic anomaly features. Next, the consistency of the trend in association strength is analyzed. This trend is described by the curve of association strength value over time. The curve of the associated anomaly feature's association strength is compared with that of the standard associated anomaly feature, analyzing the consistency of the order and duration of the rising, falling, and stable segments. The degree of agreement in the association relationship is calculated based on the consistency of the association direction and the trend of association strength. A high degree of agreement is achieved when both the association direction and the trend reach a preset threshold; otherwise, a low degree of agreement is achieved. The consistency of the association direction is judged by a binary value: 1 for consistency and 0 for inconsistency. The consistency of the association strength trend is measured by calculating the dynamic time warp distance between the two curves; the smaller the dynamic time warp distance, the higher the trend consistency. This distance is converted into a consistency index between 0 and 1. The degree of agreement of the association relationship is the product of the association direction consistency result and the trend consistency index. Simultaneously, similarity is calculated for the time series variation patterns of the association features. These time series variation patterns include periodicity, trend characteristics, and the distribution of abrupt change points. Periodicity analysis transforms the time series to the frequency domain using Fourier transform, comparing the consistency of the distribution of major frequency components. Trend analysis performs linear fitting on the time series, comparing the slope and intercept of the fitted line to see if they are within a preset error range. Abrupt change point distribution analysis calculates the first difference of the time series, identifying points where the absolute value of the difference exceeds a preset threshold as abrupt change points, and comparing the number and timing of these abrupt change points. The similarity calculation of time series variation patterns involves weighted averaging of the above-mentioned periodic characteristic similarity, trend characteristic similarity, and abrupt change point distribution similarity to obtain the correlation feature similarity. The weight of each part is determined based on the importance of its influence on the correlation relationship in historical data.

[0040] Step S1234: The basic feature similarity and the associated feature similarity are weighted and summed according to a preset weight to obtain the comprehensive matching degree between the current abnormal feature and the standard abnormal feature set. The preset weight is adjusted according to the historical anomaly identification accuracy.

[0041] The preset weights are stored in the configuration file of the pattern matching unit. The initial preset weights are set based on expert experience, with the sum of the weights for basic feature similarity and related feature similarity being 1. During system operation, the preset weights are dynamically adjusted based on historical anomaly identification accuracy. The adjustment process involves periodically (e.g., weekly) calculating the anomaly identification accuracy using the current weight settings, i.e., the ratio of correctly identified anomaly cases to the total number of identified anomaly cases. When the accuracy falls below a preset threshold, a weight adjustment mechanism is activated. This mechanism uses a gradient ascent method, making minor adjustments to the weights of basic feature similarity and related feature similarity, calculating the change in accuracy after the adjustment. If the accuracy increases, the adjustment direction is retained and the adjustment continues; if the accuracy decreases, the adjustment is reversed until the weight combination that maximizes the accuracy is found, and this weight combination is updated in the configuration file. When calculating the overall matching degree, the pattern matching unit reads the current preset weights from the configuration file, then multiplies the basic feature similarity by the basic feature weight and the related feature similarity by the related feature weight. The sum of these two is the overall matching degree between the current anomaly feature and the standard anomaly feature set.

[0042] Step S1235: If multiple sets of standard anomaly features have the same overall matching degree and all have the highest value, then further extract the key features from the basic anomaly features and the associated anomaly features, perform a second comparison between the key features and the key standard features in each set of standard anomaly features, and determine a unique set of standard anomaly features based on the second comparison results. The key features are the features that have the greatest impact on service anomalies.

[0043] After calculating the overall matching degree of all standard anomaly feature sets, the pattern matching unit sorts the overall matching degrees and finds the maximum value. If only one standard anomaly feature set has the maximum overall matching degree, it is directly used as the matching result. If multiple standard anomaly feature sets have the same overall matching degree and are all the highest values, a secondary comparison is required. The core of the secondary comparison is the extraction of key features. The extraction of key features is based on a feature importance evaluation algorithm, which determines the importance of features based on the contribution of each feature to the occurrence of the anomaly in historical anomaly cases. The contribution is measured by calculating the frequency of occurrence of the feature when the anomaly occurs and the mutual information value with the anomaly result. The higher the frequency of occurrence and the larger the mutual information value, the higher the importance of the feature. The top-ranked features in terms of importance are selected from the basic anomaly features and related anomaly features as key features. The number of selected features is determined according to the complexity of the anomaly type, usually 3 to 5. For each standard anomaly feature set with the highest overall matching degree, the corresponding key standard features are extracted from its standard basic anomaly features and standard related anomaly features. Then, the extracted key features are compared a second time with each key standard feature. The method for the second comparison is similar to that for the basic feature comparison and the related feature comparison, but the comparison process is more refined. For example, higher-order statistics (such as skewness and kurtosis) are used for comparison when calculating the similarity of numerical distributions, and the analysis of the lag between related features is added when comparing related relationships. Based on the results of the second comparison, the matching score of the key features is calculated, and the set of standard abnormal features with the highest score is determined as the unique matching result.

[0044] Step S124: Calculate the matching degree between the current abnormal feature and each set of standard abnormal features based on the comparison results. Determine the corresponding abnormal type identifier based on the set of standard abnormal features with the highest matching degree. The abnormal type identifier is used to distinguish different types of service abnormalities. At the same time, extract features directly related to the abnormal type identifier from the basic abnormal features and the associated abnormal features as abnormal association features.

[0045] During the comparison process, for each set of standard anomaly features, the matching degree between the basic anomaly features and the standard basic anomaly features, as well as the matching degree between the associated anomaly features and the standard associated anomaly features, are calculated. The calculation of the basic anomaly feature matching degree considers factors such as the feature type and the degree of deviation in value; the calculation of the associated anomaly feature matching degree considers factors such as the type of association and the degree of deviation in the standardized association strength value. The basic anomaly feature matching degree and the associated anomaly feature matching degree are combined according to a certain weight ratio to obtain the overall matching degree between the current anomaly feature and the set of standard anomaly features. All sets of standard anomaly features in the anomaly pattern library are traversed to find the set of standard anomaly features with the highest overall matching degree. The anomaly type corresponding to this set of standard anomaly features is the possible anomaly type for the current service, and its corresponding identifier is the anomaly type identifier. For example, if the set of standard anomaly features with the highest matching degree corresponds to "insufficient memory resources causing service response delay anomaly," then the anomaly type identifier is this type. The extraction of anomaly associated features involves filtering features directly related to the determined anomaly type identifier from the basic anomaly features and associated anomaly features. For example, for the type "Insufficient memory resources leading to abnormal service response delay", basic abnormal features related to memory usage are extracted from basic abnormal features, and the correlation abnormal features between memory usage and process response delay are extracted from the associated abnormal features, which together constitute abnormal correlation features.

[0046] Step S125: Integrate the anomaly type identifier and the anomaly association features according to a preset format to generate a service anomaly pattern recognition result.

[0047] The preset format specifies the structure and content organization of the service anomaly pattern recognition results, including the field names and data types of the anomaly type identifier, the order of anomaly-related features, and feature descriptions. During integration, the identified anomaly type identifiers and extracted anomaly-related features are filled and organized according to the preset format to form a structured service anomaly pattern recognition result. For example, the anomaly type identifier field is explicitly marked as "Insufficient memory resources causing service response delay anomaly," and the anomaly-related feature section lists, in turn, a detailed description of the basic memory usage anomaly feature, a detailed description of the memory usage and process response delay related anomaly feature, etc. The generated service anomaly pattern recognition results will serve as the basis for subsequent service adaptation model parameter adjustments.

[0048] Step S130: Based on the service anomaly pattern recognition result, call the corresponding service adaptation model, perform parameter adjustment processing on the service adaptation model, and generate a service stability control strategy. The service stability control strategy includes a resource allocation scheme and task scheduling rules.

[0049] Based on the anomaly type identified in the service anomaly pattern recognition results, a service adaptation model specifically designed for that anomaly type is retrieved from the model library. This service adaptation model is trained using historical optimization experience and machine learning algorithms. It adjusts model parameters according to the input anomaly correlation features, thereby generating effective service stability control strategies to resolve current service anomalies and ensure stable service operation.

[0050] Step S131: Parse the anomaly type identifier in the service anomaly pattern recognition result. According to the preset correspondence between anomaly type and service adaptation model, call the service adaptation model that matches the anomaly type identifier from the model library. Different anomaly type identifiers correspond to different service adaptation models. Each service adaptation model has optimization and control capabilities for specific types of service anomalies.

[0051] The pre-defined mapping between exception types and service adaptation models is stored in the model management module. This mapping specifies the name, version number, storage path, and other information of the service adaptation model corresponding to each exception type identifier. After parsing the service exception pattern recognition results to obtain the exception type identifier, the model management module uses this identifier to find the corresponding service adaptation model information and then calls the appropriate service adaptation model from the model library according to the storage path. For example, when the exception type identifier is "Insufficient memory resources causing service response delay exception," the "Memory Resource Optimization Adaptation Model" is called according to the mapping.

[0052] Step S132: Extract the abnormal association features from the service abnormality pattern recognition results, classify the abnormal association features by attribute, determine the features reflecting resource problems, process problems, and task problems, and input the classified abnormal association features into the called service adaptation model as the core basis for adjusting the service adaptation model parameters.

[0053] Anomaly correlation features encompass multiple attributes, and classifying these attributes allows for a clearer analysis of the causes of anomalies. Features reflecting resource issues include anomalies related to resource consumption such as memory, processor, and network bandwidth, such as memory usage exceeding normal ranges or excessively high processor utilization. Features reflecting process issues include anomalies related to the running status of service processes, such as abnormal process response delays or process interaction failures. Features reflecting task issues include anomalies related to task processing, such as excessively long task completion times or a sudden increase in the number of tasks received. After classification, the above-mentioned anomaly correlation features of different categories are organized according to the input format required by the service adaptation model and then input into the invoked service adaptation model. Upon receiving these features, the service adaptation model uses them as the core basis for analyzing the current anomaly situation and adjusting model parameters.

[0054] Step S133: After receiving the classified anomaly association features, the service adaptation model reads its own initial parameter configuration, which includes resource allocation parameters, task scheduling parameters, and process control parameters. Based on the influence of different types of features in the anomaly association features, the corresponding initial parameters are adjusted.

[0055] The initial parameter configuration of the service adaptation model is a set of preset parameter values ​​after model training. These parameters control the model's behavior when no adjustments are made. Resource allocation parameters include memory allocation ratio coefficients, processor time slice allocation weights, and network bandwidth allocation priorities; task scheduling parameters include task priority determination coefficients, task allocation strategy parameters, and task queue management parameters; process control parameters include process start thresholds, process restart condition parameters, and process resource limit parameters. After receiving the categorized anomaly correlation features, the service adaptation model analyzes the degree of influence of different types of anomaly correlation features on service anomalies using a built-in feature impact assessment algorithm. For example, for "insufficient memory resources causing abnormal service response delays," anomaly correlation features reflecting resource problems (such as abnormal memory usage features) have the highest impact. Based on the degree of influence of each type of feature, the model adjusts the corresponding parameters in the initial parameter configuration. For example, for memory resource problems, the weight value corresponding to the high-priority conversion task in the memory allocation ratio coefficient is increased, and resource allocation parameters related to memory allocation are adjusted.

[0056] Step S134: Run the service adaptation model with adjusted parameters. The service adaptation model analyzes the current operational bottleneck of the document conversion service and generates a resource allocation scheme based on the adjusted parameters. The resource allocation scheme includes the memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio of different service processes. At the same time, the service adaptation model performs priority analysis on the conversion tasks to be processed and generates task scheduling rules. The task scheduling rules include the task priority sorting method and the method of assigning tasks to different service processes.

[0057] Step S1341: Run the service adaptation model with adjusted parameters, obtain the current total system resource data of the document conversion service through the data interface. The total system resource data includes the total memory capacity, the total number of processor cores and the total network bandwidth. At the same time, obtain the current resource usage data of each service process, compare the total system resource data with the current resource usage data, determine the remaining resources, and analyze the bottleneck of the current resource allocation in combination with the adjusted resource allocation parameters.

[0058] After the service adaptation model with adjusted parameters starts, it first initializes its internal data buffer and computing module. It establishes a connection with the resource monitoring module of the document conversion service through a preset data interface, using the TCP / IP protocol and JSON format. The service adaptation model sends a resource data request command to the resource monitoring module, containing the requested resource type (total memory capacity, total processor cores, total network bandwidth) and a timestamp. Upon receiving the request, the resource monitoring module collects the current total system resource data. Total memory capacity is obtained by reading the server operating system's memory management interface, which returns the total available memory after subtracting the system kernel's reserved memory from the total physical memory. The total number of processor cores is obtained by reading the CPU's hardware information, including both physical and logical cores; here, the logical cores are used as the total number of processor cores. Total network bandwidth is determined by testing the server's network card's maximum transmission rate and considering the bandwidth limitations in the current network topology. Simultaneously, the resource monitoring module collects the current resource usage data of each service process, including memory usage, processor utilization (in units of cores), and network bandwidth usage for each process. The system's total resource data and current resource usage data are sent to the service adaptation model. Upon receiving them, the model stores both in a data buffer. Then, the remaining resource amount is calculated, which equals the total system resources minus the current resource usage, yielding the remaining memory capacity, remaining processor cores, and remaining network bandwidth. Combined with adjusted resource allocation parameters, such as memory allocation elasticity coefficients, processor allocation priority weights, and network bandwidth guarantee thresholds, the remaining resource amount and the current resource requirements of each process are analyzed. The analysis process employs a decision tree algorithm, using the remaining resource amount, historical resource requirements of each process, and adjusted resource allocation parameters as input features. Through branching in the decision tree, the bottleneck of the current resource allocation is determined. For example, if the remaining memory capacity is lower than the sum of the adjusted memory allocation elasticity coefficient and the basic memory requirements of each process, memory resources are identified as the bottleneck; if the remaining processor cores cannot meet the processor requirements of high-priority processes, processor resources are identified as the bottleneck; and if the remaining network bandwidth is lower than the guarantee threshold, network bandwidth resources are identified as the bottleneck.

[0059] Step S1342: Based on the resource bottleneck analysis results, the adjusted service adaptation model calculates the resource requirement priority of each service process according to the predefined service process weight coefficient and the standardized real-time monitoring task urgency index; the service process weight coefficient is determined based on its preset importance level, and the task urgency index is determined based on the delay threshold and deadline of the task queue and is standardized.

[0060] The service adaptation model predefines a service process weight coefficient table, assigning a weight coefficient to each service process. The weight coefficient ranges from 0 to 1. The importance level of a service process is determined by the impact of the task type it handles on the enterprise's business. This impact is jointly assessed by the business and IT departments and categorized into three levels: core, important, and general. The core level corresponds to the highest weight coefficient, and the general level corresponds to the lowest. The weight coefficient table is stored in the service adaptation model's configuration file and can be manually updated based on business adjustments. Real-time monitoring of the task urgency metric is provided by the task scheduling module. This metric is calculated for each service process's currently processing task queue. The task urgency metric is calculated based on the latency threshold and deadline of all tasks in the task queue. The latency threshold is the maximum allowed waiting time for a task, and the deadline is the final time a task must complete. For each task, the ratio of its remaining time (deadline minus current time) to the latency threshold is calculated; the smaller the ratio, the higher the task urgency. The urgency index of all tasks in the task queue is weighted and averaged to obtain the task urgency index of the service process. The weights are the proportion of the task's data volume to the total data volume in the queue. The obtained task urgency index is then standardized by dividing the index value by the historical maximum task urgency index value, resulting in a standardized task urgency index ranging from 0 to 1. When calculating the resource requirement priority of each service process, the service adaptation model multiplies the service process weight coefficient by the standardized task urgency index to obtain a preliminary priority score. Then, the preliminary priority score is adjusted according to the resource bottleneck type. For example, if the bottleneck is memory resources, an adjustment factor is added to the preliminary priority score of memory-intensive service processes. The adjustment factor is determined based on the process's memory requirement coefficient; the higher the memory requirement coefficient, the larger the adjustment factor. The adjusted score is the resource requirement priority of each service process.

[0061] Step S1343: Based on the priority of the resource requirements, allocate the remaining resources proportionally to each service process to generate a resource allocation scheme that includes the memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio for different service processes.

[0062] The service adaptation model prioritizes the resource requirements of each service process from highest to lowest, forming a resource allocation sequence. For the remaining memory capacity, memory resources are allocated to each service process sequentially according to this sequence. During allocation, the basic memory requirements of each service process are satisfied first. These basic requirements are determined based on the process type and its historical minimum memory usage, and are stored in the service process configuration table. After the basic memory requirements are met, the remaining memory resources are allocated proportionally to the resource requirement priority score of each service process relative to the total priority score (the sum of the priority scores of all processes). For example, if a service process has a priority score of P and a total priority score of SumP, then the additional memory allocated to this process is the remaining memory capacity multiplied by P / SumP. The total memory allocation for this process is the sum of the basic memory requirement and the additional memory allocation, which is then used to calculate its memory allocation ratio (total memory allocation divided by the total system memory capacity). Similarly, the allocation of remaining processor cores prioritizes the basic processor core requirements of each service process. These basic processor core requirements are determined based on the process's concurrent processing capabilities. For example, a text conversion process requires 1 core, and a table conversion process requires 2 cores. Once the basic requirements are met, the remaining processor cores are allocated according to resource demand priority, calculated similarly to memory allocation, resulting in the processor allocation ratio for each process (the number of allocated cores divided by the total number of processor cores). Network bandwidth allocation first ensures the basic network bandwidth requirements of each service process, determined by the average data transfer rate during task processing. Then, the remaining network bandwidth is allocated according to priority, resulting in the network bandwidth allocation ratio for each process (allocated bandwidth divided by the total network bandwidth). The calculated memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio are organized by service process name to form a resource allocation scheme. This scheme is in XML format and includes process identifiers, resource allocation ratios, and explanations of the allocation criteria.

[0063] Step S1344: The service adaptation model simultaneously obtains the task information of all currently pending transformation tasks. The task information includes the document type, document data volume, task submission time, and task recipient requirements. The processing difficulty is determined based on the document type in the task information, the processing time is determined based on the document data volume, and the urgency is determined based on the task submission time and recipient requirements.

[0064] The service adaptation model obtains task information for all currently pending conversion tasks through a data interface with the task queue management module. The data interface uses a RESTful style, with the service adaptation model sending an HTTP GET request. The request parameters include the task queue identifier and status (pending). The task queue management module returns a JSON-formatted task list. Each task object in the list contains a task ID, document type (e.g., .docx, .xlsx, .pptx), document data size (in bytes), task submission time (timestamp format), and task recipient requirements (including urgency, output format requirements, and encryption requirements). The service adaptation model parses the task list, extracts the task information for each task, and stores it in a task information table. When determining the processing difficulty based on the document type, it queries a preset document type difficulty coefficient table. Each document type is assigned a difficulty coefficient based on factors such as the number of computational steps required during the conversion process, algorithm complexity, and dependence on external libraries. For example, presentation documents (.pptx) have a higher difficulty coefficient than plain text documents (.docx) because they contain images and animations. Processing time is determined based on document data volume. A mapping model between document data volume and processing time is established, trained using historical task data. The inputs are document data volume and document type difficulty coefficient, with the output being the predicted processing time. The model employs a multiple linear regression algorithm. A positive coefficient for document data volume indicates that larger data volumes result in longer processing times, and a positive coefficient for the difficulty coefficient also indicates that higher difficulty levels lead to longer processing times. Urgency is determined based on task submission time and recipient requirements. The later the task submission time (closer to the current time), the higher the urgency. For tasks marked "urgent" in the recipient's requirements, the urgency coefficient is increased by a preset value, determined according to the company's internal task priority rules. For example, the urgency coefficient for urgent tasks is increased by 0.3 (the urgency coefficient ranges from 0 to 1).

[0065] Step S1345: Based on processing difficulty, processing time and urgency, the service adaptation model establishes a task priority evaluation system, scores the priority of each conversion task to be processed, and determines the task priority sorting method according to the score from high to low, forming the priority sorting part in the task scheduling rules.

[0066] The task priority evaluation system includes evaluation indicators, indicator weights, and a scoring calculation method. Evaluation indicators include processing difficulty, processing time, and urgency, with each indicator's value standardized to between 0 and 1. Indicator weights are determined using the analytic hierarchy process (AHP). An evaluation team composed of enterprise IT operations experts and business department representatives compares the importance of each indicator pairwise, constructing a judgment matrix. The weight of each indicator is obtained by calculating the largest eigenvalue and its corresponding eigenvector of the judgment matrix, with urgency typically having a higher weight than processing difficulty and processing time. The scoring calculation method multiplies the processing difficulty indicator value by its weight, the processing time indicator value by its weight, and the urgency indicator value by its weight, then the sum of these three is the task priority score. For example, if a task has a processing difficulty indicator value of D and a weight of Wd; a processing time indicator value of T and a weight of Wt; and an urgency indicator value of U and a weight of Wu, then the priority score is: The service adaptation model calculates a priority score for each of the currently pending transformation tasks, and then sorts the tasks according to their scores from highest to lowest. The sorting result is the task priority sorting method, which is recorded in the task scheduling rules as a list of task IDs as the priority sorting part.

[0067] Step S1346: Calculate the comprehensive processing capability score of each service process based on its historical processing efficiency, real-time resource utilization, and supported document types.

[0068] The service adaptation model retrieves historical processing efficiency data for each service process from the historical task processing database. Historical processing efficiency is measured by the number of tasks processed per unit time and the average task processing accuracy. The number of tasks processed per unit time is the average of the most recent period (e.g., the past 24 hours), and the average task processing accuracy is obtained by dividing the number of successfully processed tasks in the historical task database by the total number of processed tasks. Real-time resource utilization is obtained from the resource monitoring module, including current memory utilization (used memory / allocated memory), processor utilization (used cores / allocated cores), and network bandwidth utilization (used bandwidth / allocated bandwidth). The average of these three is taken as the comprehensive real-time resource utilization. The document types supported by the service process are recorded in the service process capability table. Each process supports multiple document types, represented by binary vectors. Supported document types are represented by a 1, and unsupported document types by a 0. When calculating the comprehensive processing capability score, historical processing efficiency is first converted into an efficiency score: Efficiency Score = (Number of tasks processed per unit time / Historical maximum number of tasks processed per unit time) 0.6 + Average Task Processing Accuracy 0.4, with weights set according to the degree of impact on processing capability. Next, the resource sufficiency score is calculated: Resource sufficiency score = 1 - Overall real-time resource utilization rate. The lower the resource utilization rate, the higher the resource sufficiency score. Then, the document type matching score is calculated. For a given document type in the current pending tasks, if the service process supports that type, the matching score is 1; otherwise, it is 0. If there are multiple pending tasks, the average matching score is taken. The overall processing capacity score = Efficiency score 0.5 + Resource sufficiency score 0.3 + Document type matching score The weights can be adjusted according to the current optimization goals of the system. For example, when the system focuses on processing speed, the weight of the efficiency score can be increased.

[0069] Step S1347: Set the resource availability threshold and the processing capacity score threshold.

[0070] Resource availability thresholds include memory availability threshold, processor availability threshold, and network bandwidth availability threshold. These represent the minimum available proportion of memory, processor, and network bandwidth resources for a service process to receive new tasks. These thresholds are determined based on the type of service process and its historical stable resource availability. For example, text conversion processes have lower memory dependency, so their memory availability threshold can be set to a lower value, while table conversion processes have a higher memory availability threshold. The processing capacity score threshold is the minimum score required to determine whether a service process can handle new tasks. This threshold is determined through statistical analysis of the comprehensive processing capacity scores of service processes when they have been able to stably handle tasks historically. The lower limit of the historical scores is used as the threshold; service processes below this threshold are not assigned new tasks. The resource availability threshold and processing capacity score threshold are stored in the parameter configuration file of the service adaptation model and can be manually adjusted by the system administrator based on actual operating conditions. Adjustment records are automatically saved in the audit log.

[0071] Step S1348: Assign tasks whose priority meets the preset sorting criteria to service processes whose comprehensive processing capability score is higher than the processing capability score threshold and whose real-time resource occupancy rate is lower than the resource availability rate threshold, thereby forming the task allocation method part of the task scheduling rules and completing the generation of resource allocation scheme and task scheduling rules.

[0072] The preset sorting criterion refers to a task priority score higher than a certain preset value. This preset value is determined based on the length of the current task queue and the average priority score. When the task queue is long, the preset value can be appropriately increased to prioritize more urgent tasks. The service adaptation model traverses the task list sorted by priority. For tasks whose priority meets the preset sorting criterion, its document type is obtained. Then, all service processes are traversed, and service processes with a comprehensive processing capability score higher than the processing capability score threshold and a real-time resource utilization rate lower than the resource availability rate threshold are selected to form a candidate service process list. In the candidate service process list, the most suitable service process is selected based on the matching degree between the document types supported by the service process and the document type of the current task, the comprehensive processing capability score of the service process, and the matching degree with the task priority (high-priority tasks are assigned to processes with higher scores). The selection method adopts a weighted voting system, assigning weights to the matching degree, score, and priority matching degree of each candidate process, calculating a comprehensive score, and selecting the process with the highest score. The task is assigned to the selected service process, and the correspondence between the task ID and the service process ID is recorded to form the task allocation method section. The resource allocation scheme and task scheduling rules (including priority sorting and task allocation method) are integrated into a JSON document, and the scheme generation timestamp and version number are added to complete the generation of service stability control strategy.

[0073] Step S135: Perform compatibility verification on the generated resource allocation scheme and task scheduling rules to ensure that the total resource ratio of each process in the resource allocation scheme meets the system's total resource limit, and that the task allocation method in the task scheduling rules matches the processing capacity of the service process. After the verification is passed, integrate the two to generate a service stability control strategy.

[0074] Compatibility verification is a crucial step in ensuring that resource allocation schemes and task scheduling rules can be effectively executed in a real-world system. For resource allocation schemes, the sum of the memory allocation ratios for each service process is verified to match the total available memory ratio of the system, the sum of the processor allocation ratios is verified to match the total available processor resources of the system, and the sum of the network bandwidth allocation ratios is verified to match the total available network bandwidth of the system. This ensures that the resource allocation scheme complies with the total system resource limits and avoids resource allocation conflicts. For task scheduling rules, the task allocation method is verified to match the processing capacity of the service processes. For example, if the processing capacity limit of a service process is to handle a certain number of specific types of tasks simultaneously, the number and types of tasks allocated to that process in the task scheduling rules must be within its processing capacity to avoid task failures or service crashes due to task allocation exceeding the process's processing capacity. If incompatibility issues are found in the resource allocation scheme or task scheduling rules during verification, the process returns to the service adaptation model for parameter adjustment and scheme generation until the compatibility verification is passed. After successful verification, the resource allocation scheme and task scheduling rules are integrated according to a preset format to form a complete service stability control strategy. This service stability control strategy includes all the resource allocation and task scheduling instructions required for stable service operation.

[0075] Step S140: Input the service stability control strategy into the execution control module of the document conversion service, execute the service stability control strategy, and obtain the service operation parameter adjustment result.

[0076] After the service stability control strategy is generated, it needs to be passed to the execution control module of the document conversion service. The execution control module is responsible for adjusting the service's operating parameters according to the instructions in the strategy. The execution control module is the bridge connecting strategy generation and actual service operation adjustment. By executing the control strategy, the service's operating status is improved, and previously existing anomalies are resolved.

[0077] Step S141: The service stability control strategy is input into the execution control module of the document conversion service through the data transmission channel. The execution control module has a built-in strategy parsing unit. The strategy parsing unit performs structured parsing on the service stability control strategy and extracts the resource allocation scheme and task scheduling rules therein.

[0078] The data transmission channel employs encrypted transmission to ensure the security and integrity of the service stability control strategy during transmission. Upon receiving the control strategy, the execution control module passes it to the built-in strategy parsing unit. The strategy parsing unit performs structured parsing according to the preset format of the control strategy, separating it into two parts: a resource allocation scheme and task scheduling rules. When parsing the resource allocation scheme, specific parameters such as the memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio for each service process are extracted. When parsing the task scheduling rules, the specific rules for task priority ranking (such as ranking priority based on factors like task submission time, urgency, and document type) and the specific methods for assigning tasks to different service processes (such as allocation based on process load, processing capacity, and historical success rate) are extracted. After parsing, the extracted information is stored in the execution control module's temporary data storage area.

[0079] Step S142: Based on the parsed resource allocation scheme, the execution control module sends a resource adjustment instruction to the resource allocation module of the document conversion service. The resource adjustment instruction includes the target memory usage, target processor usage, and target network bandwidth usage of each service process. After receiving the instruction, the resource allocation module gradually adjusts the resource usage of each service process so that the final resource allocation ratio conforms to the resource allocation scheme.

[0080] The execution control module calculates the target memory usage, target processor usage, and target network bandwidth usage for each service process based on the parsed resource allocation scheme and the current total system resources. For example, it calculates the target memory usage for each process based on the memory allocation ratio and the total system memory capacity. Then, it generates resource adjustment instructions containing these target usages and sends them to the resource allocation module. The resource allocation module is responsible for managing the allocation of server hardware resources. Upon receiving the resource adjustment instructions, it adjusts the resource usage of each service process in a gradual manner. During the gradual adjustment process, the resource allocation module monitors the resource usage and service running status of each process in real time to avoid service process interruptions or task failures due to excessively rapid resource adjustments. For example, for service processes that need increased memory allocation, their memory quota is increased gradually in multiple steps, and the stability of process operation is observed after each adjustment until the target memory usage is reached; for service processes that need reduced memory allocation, some memory resources are gradually released while ensuring that their current processing tasks are not affected. Through the above methods, the resource usage ratio of each service process is ultimately made consistent with the provisions in the resource allocation scheme.

[0081] Step S143: Based on the parsed task scheduling rules, the execution control module sends a task adjustment instruction to the task scheduling module of the document conversion service. The task adjustment instruction contains a task priority sorting list and the target service process identifier corresponding to each task.

[0082] The execution control module reorders the queue of currently pending conversion tasks according to the parsed task scheduling rules, generating a task priority sorting list. This list sorts all pending tasks from highest to lowest priority, as specified in the task scheduling rules. Simultaneously, based on the task allocation method in the scheduling rules and the current status of each service process (e.g., resource usage, processing capacity, number of tasks), a target service process is assigned to each task, and a corresponding target service process identifier is generated. Then, the execution control module integrates the task priority sorting list and the target service process identifiers for each task into a task adjustment instruction and sends it to the task scheduling module.

[0083] Step S144: After receiving the task adjustment instruction, the task scheduling module reorders the current queue of tasks to be processed and assigns the tasks to the corresponding service processes according to the sorting result.

[0084] Upon receiving a task adjustment instruction, the task scheduling module first reorganizes the current queue of pending tasks according to the task priority ranking list, adjusting the execution order of tasks so that high-priority tasks are placed at the front of the queue and receive processing resources first. Then, based on the target service process identifier corresponding to each task, the module assigns the task to the task processing queue of the appropriate service process. During task allocation, the task scheduling module maintains communication with each service process, obtaining real-time status of the service process's task processing queue to ensure the accuracy and timeliness of task allocation. For example, a high-urgency text document conversion task, ranking first in the task priority ranking list, is assigned to the dedicated text document conversion service process corresponding to the target service process identifier.

[0085] Step S145: During the process of resource adjustment by the resource allocation module and task adjustment by the task scheduling module, the execution control module collects the change data of service operation parameters through the real-time monitoring unit. The change data of service operation parameters includes change data of resource usage, change data of task processing efficiency, and change data of process response. The collection frequency is adapted to the change rate of service operation parameters.

[0086] Step S1451: The real-time monitoring unit of the execution control module determines the acquisition frequency. The acquisition frequency is determined based on the rate of change of historical service operation parameters. Parameters with a fast rate of change are acquired at a more frequent acquisition frequency, while parameters with a slow rate of change are acquired at a relatively gentler acquisition frequency, so that the acquisition frequency is adapted to the rate of change of the parameters.

[0087] When the real-time monitoring unit starts, it first retrieves historical change records of various service operation parameters from the historical monitoring database. These records include the parameter name, sampling timestamp, and parameter value. For each parameter, its historical change rate is calculated. The change rate is measured by the average absolute change of the parameter value per unit time, with the unit time being 10 times the parameter sampling period. For example, if the sampling period for a parameter is T, and the parameter value changes from V1 to V2 within a time window of 10T, then the change rate is |V2-V1| / 10T. Based on the magnitude of the historical change rate, parameters are categorized into three types: fast-changing parameters, medium-changing parameters, and slow-changing parameters. Fast-changing parameters include processor utilization and network bandwidth utilization, which have relatively high change rates. Medium-changing parameters include memory utilization and task processing progress. Slow-changing parameters include process startup time and total task completion time. Different basic sampling frequency ranges are preset for different types of parameters. The basic sampling frequency range for fast-changing parameters is higher than that for medium-changing parameters, and vice versa. The real-time monitoring unit dynamically adjusts the actual sampling frequency within the corresponding basic sampling frequency range based on the historical change rate of the current parameter. The adjustment method compares the current rate of change with the average historical rate of change for that parameter. If the current rate of change is higher than the average, the sampling frequency is increased within the baseline range; if it is lower than the average, the sampling frequency is decreased. The adjustment range of the sampling frequency is determined based on the degree of deviation of the rate of change; the greater the deviation, the greater the adjustment, but it does not exceed the baseline sampling frequency range. The determined sampling frequency is stored in the sampling configuration table of the real-time monitoring unit, with each parameter corresponding to one configuration record, including the parameter ID, the current sampling frequency, and the next evaluation and adjustment time.

[0088] Step S1452: During the resource adjustment process performed by the resource allocation module, the real-time monitoring unit collects the resource usage data of each service process according to the determined collection frequency. After each collection, the difference between the current resource usage data of the service process and the resource usage data collected last time is calculated. The difference is associated with the collection timestamp, and the change values ​​of different types of resources are standardized to form resource usage change data. The resource usage change data includes the standardized memory usage change value, processor usage change value, and network bandwidth usage change value.

[0089] The real-time monitoring unit periodically sends resource usage data collection requests to the resource allocation module based on the parameter ID and collection frequency in the collection configuration table. The request includes the target service process ID and the type of resource to be collected (memory, processor, network bandwidth). Upon receiving the request, the resource allocation module collects the current resource usage data of the specified service process through the system call interface provided by the operating system. Memory usage data is obtained by reading the process's virtual memory size and physical memory size, with the physical memory size used as the current memory usage data. Processor usage data is obtained by reading the process's CPU time slice usage, calculating the CPU utilization rate within the collection interval, and then multiplying it by the allocated number of processor cores. Network bandwidth usage data is obtained by monitoring the process's network socket and counting the total number of bytes sent and received per unit time. After collection, the resource allocation module returns the current resource usage data and the collection timestamp to the real-time monitoring unit. The real-time monitoring unit calculates the difference between the current resource usage data and the previously collected resource usage data (stored in a local cache, with the service process ID and resource type as keys) to obtain the original change value (current value - previous value). The original change values ​​are standardized by dividing them by the maximum change magnitude for that resource type during historical adjustments. The maximum change magnitude is determined by statistically analyzing the maximum absolute change value for that resource type during historical resource adjustments. The standardized change values ​​range from -1 to 1, with positive values ​​indicating an increase in resource usage and negative values ​​indicating a decrease. The standardized memory usage change values, processor usage change values, and network bandwidth usage change values ​​are then associated with the collection timestamp and service process ID and stored in a resource usage change data buffer to form resource usage change data.

[0090] Step S1453: During the task adjustment process performed by the task scheduling module, the real-time monitoring unit collects the number of tasks completed by each service process within a unit of time, calculates the difference between the current unit of time processing volume and the previous unit of time processing volume, and obtains task processing efficiency change data. At the same time, the real-time monitoring unit collects the response time from the time each service process receives a request to the time it returns a processing result, calculates the difference between the current response time and the previous response time, and obtains process response change data.

[0091] The real-time monitoring unit establishes a long-lived connection with the task scheduling module, receiving task processing status update events through a subscription mechanism. The unit time is set based on the average processing time of the tasks; for example, for short tasks (average processing time less than 1 minute), the unit time is set to 1 minute; for long tasks (average processing time greater than 10 minutes), the unit time is set to 5 minutes. At the end of each unit time, the real-time monitoring unit sends a task completion query request to the task scheduling module, containing the service process ID and the time window (current unit time). The task scheduling module queries the number of tasks marked as "completed" by the service process within the specified time window and returns it to the real-time monitoring unit. The real-time monitoring unit calculates the difference between the current unit time processing volume and the previous unit time processing volume (read from the local cache, with the key being the service process ID and the unit time identifier), obtaining the task processing efficiency change data (current processing volume - previous processing volume). For collecting process response time data, the real-time monitoring unit intercepts communication messages between the service process and the task scheduling module. When the service process receives a task request message, it records the receiving timestamp; when the service process sends a task processing result message, it records the sending timestamp. The response time is the sending timestamp minus the receiving timestamp. The real-time monitoring unit periodically (with a collection frequency consistent with the preset frequency for process response change data) extracts the most recent response time for each service process from the message log. It then calculates the difference between this time and the previously collected response time (stored in a local cache) to obtain the process response change data (current response time - previous response time). The task processing efficiency change data and the process response change data are associated with the service process ID and the unit time identifier / collection timestamp, respectively, and stored in the corresponding data buffers.

[0092] Step S1454: The real-time monitoring unit performs real-time verification on the collected resource usage change data, task processing efficiency change data, and process response change data. The verified resource usage change data, task processing efficiency change data, and process response change data are stored according to the service process, with each service process corresponding to a complete set of change data.

[0093] The verification module of the real-time monitoring unit first verifies the resource usage change data. Verification includes data integrity (whether it includes the service process ID, timestamp, and three resource change values), data rationality (whether the standardized change value is between -1 and 1), and timestamp continuity (whether the current timestamp is later than the previous timestamp and the interval is within a reasonable range, determined based on the collection frequency). For task processing efficiency change data, verification includes whether the number of tasks is a non-negative integer, whether the unit time identifier is correct, and whether the change value is within a preset reasonable range (determined based on the maximum and minimum values ​​of historical task processing efficiency changes). For process response change data, verification includes whether the response time is a non-negative value and whether the change value is within a preset range (to avoid misjudgments due to outliers). During the verification process, if incomplete or unreasonable data is found, the real-time monitoring unit sends a re-collection request to the resource allocation module or task scheduling module, retrying a maximum of a preset number of times (e.g., 3 times). If the verification still fails after retrying, the data is marked as invalid, and an error log is recorded. Data that passes verification is sent to the categorized storage module. The categorized storage module creates data directories based on the service process ID, and within each directory, it creates three subdirectories: "Resource Usage Changes," "Task Processing Efficiency Changes," and "Process Response Changes." The corresponding data is stored as JSON files in the respective subdirectories, with filenames including the timestamp of data generation for easy retrieval in chronological order. Within each service process's directory, these three types of change data files form a complete record of the service's operational parameter changes for that process.

[0094] Step S146: The execution control module performs time-series processing on the collected service operation parameter change data, sorts the service operation parameter change data according to the collection time order, and marks the attributes of the valid data, marking the corresponding service process, data type and collection time. After processing, the service operation parameter adjustment result is formed, which reflects the change in service operation status after the service stability control strategy is implemented.

[0095] After collecting data on changes in service operating parameters, the execution control module sorts the data according to the time sequence of collection to construct a time-series dataset. Then, it performs a validity check on each data point in the time series, removing invalid data due to collection errors or interference. For valid data, it adds attribute annotations, specifying the corresponding service process (e.g., "text document conversion process," "table document conversion process"), data type (e.g., "memory usage change data," "processor usage change data"), and specific collection time. The dataset, after time-series processing and attribute annotation, forms the service operating parameter adjustment result.

[0096] Step S150: Update the training sample set of the service adaptation model based on the service operation parameter adjustment results, perform iterative optimization processing on the service adaptation model, and generate an optimized service adaptation model.

[0097] The results of service operation parameter adjustments reflect the effectiveness of the control strategies generated by the service adaptation model in practical applications. By feeding these results back into the training process of the service adaptation model and iteratively optimizing the model, the model's ability to control service anomalies can be continuously improved, enabling the service adaptation model to better adapt to various complex situations during service operation.

[0098] Step S151: Analyze the service operation parameter adjustment results, extract the resource usage change data, task processing efficiency change data, and process response change data, and simultaneously retrieve the service anomaly pattern identification results and service stability control strategies corresponding to this stability control process. Associate the resource usage change data, the task processing efficiency change data, the process response change data, the service anomaly pattern identification results, and the service stability control strategies with the timestamp of the control process.

[0099] First, the results of service operation parameter adjustments are analyzed. Resource usage change data (such as sequences of changes in memory usage and processor usage for each service process), task processing efficiency change data (such as sequences of changes in the number of tasks processed per unit time and sequences of changes in task processing progress rate), and process response change data (such as sequences of changes in process response latency and sequences of changes in process interaction success rate) are extracted from the organized time-series data. Then, using the timestamps of the adjustment process recorded in the system logs, the corresponding service anomaly pattern recognition results (including anomaly type identifiers and anomaly association features) and service stability adjustment strategies (including resource allocation schemes and task scheduling rules) are retrieved during this stability adjustment process. Using the timestamps of the adjustment process as a common identifier, the extracted change data are associated with the service anomaly pattern recognition results and service stability adjustment strategies, establishing a complete data chain of "anomaly identification - strategy generation - execution effect." For example, resource usage change data within a certain time period is associated with the anomaly association features and the executed resource allocation scheme portion of the service anomaly pattern recognition results for that time period, forming a complete adjustment case dataset.

[0100] Step S152: Label the associated dataset as a new training sample. The new training sample contains input features and output labels. The input features are the abnormal association features in the service anomaly pattern recognition results. The output labels are the service stability control strategy and the corresponding service operation parameter adjustment results. Add the new training sample to the training sample set of the service adaptation model and update the content of the training sample set.

[0101] The associated dataset is labeled with training samples, defining the input features and output labels. Input features are defined as the anomaly correlation features from the service anomaly pattern recognition results; these features describe the specific manifestations and related factors of service anomalies. Output labels are defined as the service stability control strategy generated during this control process and the adjusted service operation parameters obtained after implementing the strategy. The output labels reflect the control measures taken for specific anomaly correlation features and their actual effects. The labeled dataset is then added as new training samples to the training sample set of the service adaptation model. The training sample set is a database storing a large amount of historical control case data. The addition of new samples expands the size and coverage of the training sample set, including more up-to-date service anomaly situations and control experiences.

[0102] Step S153: Select samples from the updated training sample set as verification samples according to a preset ratio, and use the remaining samples as training samples. The preset ratio is determined according to the total number of training samples, so that the verification samples can cover samples corresponding to different types of service anomalies.

[0103] The predetermined ratio is determined by considering the total number of training samples. When the total number of samples is large, the predetermined ratio can be appropriately reduced to ensure a sufficient number of training samples for model training; when the total number of samples is small, the predetermined ratio can be appropriately increased to ensure that the validation samples can fully reflect different types of service anomalies. When selecting validation samples, a stratified sampling method is used. The training sample set is stratified according to the type of service anomaly. A certain number of samples are selected from the samples corresponding to each anomaly type according to a predetermined ratio as validation samples, and the remaining samples are used as training samples. This method ensures that the validation samples cover samples corresponding to all different types of service anomalies, ensuring a comprehensive evaluation of the model's ability to handle various anomalies during model validation. For example, for a training sample set containing multiple anomaly types, the same proportion of samples are selected from the samples of each anomaly type to form the validation samples, ensuring the representativeness of the validation samples.

[0104] Step S154: Divide the training samples into batches, with each batch containing a preset number of training samples. Input the training samples into the service adaptation model in batch order. The service adaptation model processes the input features in each training sample and outputs a predicted service stability control strategy.

[0105] The predetermined number of samples is determined by comprehensively considering factors such as the computational power of the service adaptation model, memory resource limitations, and training efficiency requirements. The number of samples in each batch should allow the model to process them efficiently in memory. Training samples are divided into batches, with each batch containing a predetermined number of consecutive training samples. During model training, training samples are input into the service adaptation model sequentially according to the batch order. After receiving the input features (anomaly association features) from the training samples, the service adaptation model processes the input features through its internal neural network structure. Each layer of the neural network transforms and abstracts the features layer by layer according to predetermined connection relationships and weight parameters. For example, the input layer receives anomaly association features, and after multiple hidden layers of convolution, pooling, and fully connected operations, the features are deeply extracted and fused. Finally, the output layer outputs the predicted service stability control strategy, which includes the predicted resource allocation scheme and task scheduling rules.

[0106] Step S155: Compare the predictive control strategy output by the service adaptation model with the actual service stability control strategy in the training samples, calculate the differences between the two in terms of resource allocation ratio and task scheduling rules, and generate a strategy difference value. At the same time, compare the parameter change results simulated by the service adaptation model based on the predictive control strategy with the actual service operation parameter adjustment results in the training samples. Before the comparison, standardize the parameter change results, calculate the differences between the two in terms of standardized resource consumption changes, task processing efficiency changes, and process response changes, and generate a parameter difference value.

[0107] The strategy difference value is calculated separately for the resource allocation ratios and task scheduling rules in the predictive regulation strategy and the actual service stability regulation strategy. For the resource allocation ratio, the predicted memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio of each service process are compared with the corresponding ratios in the actual strategy. For the task scheduling rules, the predicted task priority ranking method and task allocation method to service processes are compared with the actual rules. These differences are combined to generate the strategy difference value. The larger the difference value, the greater the deviation between the predicted strategy and the actual strategy. The calculation of the parameter difference value first requires standardizing the parameter change results (such as simulated resource usage changes, task processing efficiency changes, and process response change data) simulated by the service adaptation model based on the predictive regulation strategy and the actual service operation parameter adjustment results in the training samples to eliminate the influence of dimensional differences. The standardization process uses the same method as the previous feature standardization to transform the data into a uniform numerical range. Then, the differences between the standardized simulated parameter change results and the actual parameter change results are compared in various dimensions, such as the similarity of the standardized memory usage change curves and the consistency of the task processing efficiency change trends. These differences are combined to generate the parameter difference value.

[0108] Step S156: Construct a loss function based on the policy difference value and the parameter difference value, calculate the partial derivatives of the loss function with respect to each parameter of the service adaptation model using the gradient descent algorithm, adjust the parameters of the service adaptation model according to the direction of the partial derivatives, adjust the loss function value to a preset range, complete the training of one batch of samples, repeat the process until all training sample batches are processed, and complete one round of training.

[0109] The loss function is constructed by combining policy variance and parameter variance values ​​with certain weights to form a comprehensive function reflecting the model's prediction error. The weights are determined based on the degree of influence of policy variance and parameter variance on model performance. For example, if policy variance has a greater impact on service regulation, a higher weight is assigned to it. The partial derivatives of the loss function with respect to the parameters of the service adaptation model (such as connection weights and biases in a neural network) are calculated using the gradient descent algorithm. These partial derivatives reflect the direction and degree of influence of parameter changes on the loss function value. Based on the direction of the partial derivatives, the parameters of the service adaptation model are adjusted according to a preset learning rate, causing the loss function value to decrease. The learning rate affects the step size of parameter adjustment and needs to be dynamically adjusted based on changes in the loss function during training. During the training of a batch of samples, the model parameters are continuously adjusted until the loss function value decreases to a preset range or reaches the maximum number of iterations, completing the training of that batch of samples. All training sample batches are processed sequentially using the same method. After all batches are processed, one round of training of the service adaptation model is completed.

[0110] Step S157: After each round of training, the validation sample is input into the currently trained service adaptation model. The service adaptation model outputs the prediction and control strategy and the prediction parameter change results for the input features of the validation sample. The strategy difference between the prediction and control strategy and the actual service stability control strategy in the validation sample is calculated. The parameter difference between the prediction parameter change results and the actual service operation parameter adjustment results in the validation sample is calculated.

[0111] After each round of training, validation samples are used to evaluate the generalization ability and training effect of the service adaptation model. Validation samples are input into the currently trained service adaptation model using the same input format as the training samples. The model processes the input features (anomaly association features) of the validation samples and outputs a prediction and regulation strategy and the simulated changes in predicted parameters based on that strategy. Then, following the same method as during training, the strategy difference between the prediction and regulation strategy and the actual service stability regulation strategy in the validation samples is calculated, as is the parameter difference between the predicted parameter changes and the actual service operation parameter adjustments in the validation samples. These differences reflect the model's predictive performance on validation samples that were not part of the training.

[0112] Step S158: Calculate the validation accuracy of the service adaptation model based on the policy difference value and parameter difference value of the validation samples. The validation accuracy reflects the degree of agreement between the prediction results of the service adaptation model and the actual results. If the validation accuracy does not reach the preset standard, adjust the training parameters and retrain the training samples according to the batch processing until the validation accuracy of the service adaptation model reaches the preset standard, thus completing the training optimization of the service adaptation model.

[0113] The validation accuracy is calculated based on the policy and parameter differences of the validation samples. This difference is converted into an accuracy metric to measure the degree of agreement between the model's predictions and the actual results. For example, when both the policy and parameter differences are less than a preset threshold, the validation sample is considered to have predicted correctly. The validation accuracy is the ratio of the number of correctly predicted validation samples to the total number of validation samples. The preset standard is a validation accuracy threshold set according to the actual needs of service stability control. This threshold reflects the minimum requirements for the model's predictive performance. If the calculated validation accuracy does not reach the preset standard, it indicates insufficient generalization ability or inadequate training of the model. Training parameters (such as learning rate, batch size, and maximum number of training epochs) need to be adjusted, and the service adaptation model should be trained again starting from the initial batch division of training samples, following the above training process. This process of adjusting training parameters and training continues until the validation accuracy of the service adaptation model on the validation samples reaches the preset standard. At this point, the model training optimization is considered complete, and an optimized service adaptation model is generated. The optimized service adaptation model can more accurately generate effective service stability control strategies based on the service anomaly pattern recognition results, improving the stability of the document conversion service.

[0114] Figure 2 The illustration shows exemplary hardware and software components of an AI-based document conversion service stability optimization system 100 that can implement the ideas of this application, according to some embodiments of this application. For example, a processor 120 can be used in the AI-based document conversion service stability optimization system 100 and to perform the functions described in this application.

[0115] For example, the AI-based document conversion service stability optimization system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the AI-based document conversion service stability optimization system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The methods of this application can be implemented according to these program instructions. The AI-based document conversion service stability optimization system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0116] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned artificial intelligence-based document conversion service stability optimization method is implemented.

[0117] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for optimizing the stability of document conversion services based on artificial intelligence, characterized in that, The method includes: Obtain a real-time running status data set of the document conversion service, perform feature extraction processing on the real-time running status data set to obtain a service running status feature set, wherein the real-time running status data set includes service process running data, resource usage data and conversion task processing data; The pre-trained service anomaly tracing module is invoked to perform anomaly pattern recognition processing on the service running status feature set, generating service anomaly pattern recognition results, which include anomaly type identifiers and anomaly association features. Based on the service anomaly pattern recognition result, the corresponding service adaptation model is invoked, the parameters of the service adaptation model are adjusted, and a service stability control strategy is generated. The service stability control strategy includes a resource allocation scheme and task scheduling rules. The service stability control strategy is input into the execution control module of the document conversion service, and the service stability control strategy is executed to obtain the service operation parameter adjustment results; The training sample set of the service adaptation model is updated based on the service operation parameter adjustment results, and the service adaptation model is iteratively optimized to generate an optimized service adaptation model. The pre-trained service anomaly tracing module is invoked to perform anomaly pattern recognition processing on the service runtime state feature set, generating service anomaly pattern recognition results, including: The service operation status feature set is input into the feature layering unit of the service anomaly tracing module. The service operation status feature set is layered according to the data source and correlation to obtain a basic feature layer and a related feature layer. The basic feature layer contains features corresponding to a single type of data, namely, independent features corresponding to service process operation data, resource consumption data, and conversion task processing data respectively. The related feature layer contains correlation features between different types of data, namely, correlation features between service process operation data and resource consumption data, correlation features between resource consumption data and conversion task processing data, and correlation features between service process operation data and conversion task processing data. For each feature in the basic feature layer, an anomaly threshold comparison is performed. By using a preset normal range for the feature, features that exceed the normal range are selected as basic abnormal features. For each associated feature in the associated feature layer, a correlation degree analysis is performed. By calculating the standardized correlation strength value of the associated features, features whose standardized correlation strength value exceeds the normal correlation range are selected as associated abnormal features. The basic anomaly features and the associated anomaly features are input into the pattern matching unit of the service anomaly tracing module. The preset anomaly pattern library is called to compare the basic anomaly features with the standard basic anomaly features one by one, and the associated anomaly features with the standard associated anomaly features one by one. The anomaly pattern library stores a set of standard anomaly features corresponding to various historical service anomalies. Each set of standard anomaly features includes standard basic anomaly features and standard associated anomaly features. Based on the comparison results, the degree of matching between the current abnormal feature and each set of standard abnormal features is calculated. The corresponding abnormal type identifier is determined according to the set of standard abnormal features with the highest degree of matching. The abnormal type identifier is used to distinguish different types of service abnormalities. At the same time, features directly related to the abnormal type identifier are extracted from the basic abnormal features and the associated abnormal features as abnormal association features. The anomaly type identifier and the anomaly association features are integrated according to a preset format to generate a service anomaly pattern recognition result.

2. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 1, characterized in that, The process involves acquiring a real-time operational status data set of the document conversion service, performing feature extraction on the real-time operational status data set to obtain a service operational status feature set, including: Collect service process running data generated during the operation of the document conversion service. The service process running data includes process startup time data, process response delay data, and process interaction data. Collect resource usage data during the operation of the document conversion service, including memory usage data, processor usage data, and network bandwidth usage data; Collect conversion task processing data when the document conversion service processes tasks. The conversion task processing data includes task received quantity data, task processing progress data, and task completion time data. The service process running data, the resource usage data, and the conversion task processing data are integrated according to the time series to form a real-time running status data set. Feature extraction processing is performed on the real-time running status data set. The process running features, resource usage features, and task processing features are extracted by the feature extraction algorithm. The process running features correspond to the key attributes of the service process running data, the resource usage features correspond to the key attributes of the resource usage data, and the task processing features correspond to the key attributes of the conversion task processing data. After the process running characteristics, resource usage characteristics, and task processing characteristics are input into the feature fusion module for standardized processing, a combination of feature splicing and weight allocation is used for fusion processing to generate a service running status feature set with a unified dimension.

3. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 1, characterized in that, The step of inputting the basic anomaly features and the associated anomaly features into the pattern matching unit of the service anomaly tracing module, calling the preset anomaly pattern library, and comparing the basic anomaly features with standard basic anomaly features one by one, and comparing the associated anomaly features with standard associated anomaly features one by one, includes: Call the preset abnormal pattern library, read all standard abnormal feature sets stored in the abnormal pattern library, and identify each standard abnormal feature set; For each set of standard anomaly features, each feature in the set of standard anomaly features is compared with the corresponding standard basic anomaly feature in terms of feature attributes. Before the comparison, all feature values ​​are standardized, and the standardized difference value between the two in the attribute dimension is calculated. The standardized difference value is negatively correlated with the degree of matching. The degree of matching corresponding to the standardized difference value conforms to a preset association relationship. At the same time, the similarity between the numerical distribution of the standardized basic anomaly features and the numerical distribution of the standardized standard basic anomaly features is calculated to obtain the basic feature similarity. For each set of standard anomaly features, the association relationship between each feature in the set of standard anomaly features and the corresponding standard association anomaly features is compared. The consistency between the two in terms of association direction and association strength change trend is analyzed, the degree of agreement of association relationship is calculated, and the similarity of the time series change pattern of the association features is calculated to obtain the association feature similarity. The basic feature similarity and the associated feature similarity are weighted and summed according to a preset weight to obtain the comprehensive matching degree between the current abnormal feature and the standard abnormal feature set. The preset weight is adjusted according to the historical anomaly identification accuracy. If multiple sets of standard anomaly features have the same overall matching degree and all have the highest value, then the key features in the basic anomaly features and the associated anomaly features are further extracted. The key features are then compared with the key standard features in each set of standard anomaly features. Based on the results of the second comparison, a unique set of standard anomaly features is determined. The key features are those that have the greatest impact on service anomalies.

4. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 1, characterized in that, The step of invoking the corresponding service adaptation model based on the service anomaly pattern recognition result, adjusting the parameters of the service adaptation model, and generating a service stability control strategy includes: The abnormal type identifier in the service abnormality pattern recognition result is parsed. According to the preset correspondence between abnormal type and service adaptation model, the service adaptation model matching the abnormal type identifier is called from the model library. Different abnormal type identifiers correspond to different service adaptation models. Each service adaptation model has optimization and control capabilities for specific types of service abnormalities. Extract the abnormal association features from the service abnormality pattern recognition results, classify the abnormal association features by attributes, determine the features that reflect resource problems, process problems, and task problems, and input the classified abnormal association features into the called service adaptation model as the core basis for adjusting the service adaptation model parameters. After receiving the classified anomaly correlation features, the service adaptation model reads its own initial parameter configuration, which includes resource allocation parameters, task scheduling parameters, and process control parameters. Based on the influence of different types of features in the anomaly correlation features, the corresponding initial parameters are adjusted. The service adaptation model with adjusted parameters is run. The service adaptation model analyzes the current operational bottleneck of the document conversion service and generates a resource allocation scheme based on the adjusted parameters. The resource allocation scheme includes the memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio of different service processes. At the same time, the service adaptation model performs priority analysis on the conversion tasks to be processed and generates task scheduling rules. The task scheduling rules include the task priority sorting method and the method of assigning tasks to different service processes. The generated resource allocation scheme and task scheduling rules are subjected to compatibility verification to ensure that the total resource ratio of each process in the resource allocation scheme meets the system's total resource limit, and that the task allocation method in the task scheduling rules matches the processing capacity of the service process. After the verification is passed, the two are integrated to generate a service stability control strategy.

5. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 4, characterized in that, The service adaptation model, after adjusting the parameters, analyzes the current operational bottlenecks of the document conversion service and generates a resource allocation scheme based on the adjusted parameters. This resource allocation scheme includes memory allocation ratios, processor allocation ratios, and network bandwidth allocation ratios for different service processes. Simultaneously, the service adaptation model performs priority analysis on currently pending conversion tasks and generates task scheduling rules, including: Run the service adaptation model with adjusted parameters, obtain the current total system resource data of the document conversion service through the data interface. The total system resource data includes the total memory capacity, the total number of processor cores and the total network bandwidth. At the same time, obtain the current resource usage data of each service process, compare the total system resource data with the current resource usage data, determine the remaining resources, and analyze the bottleneck of the current resource allocation in combination with the adjusted resource allocation parameters. Based on the resource bottleneck analysis results, the adjusted service adaptation model calculates the resource requirement priority of each service process according to the predefined service process weight coefficient and the standardized real-time monitoring task urgency index. The service process weight coefficient is determined based on its preset importance level, and the task urgency index is determined based on the delay threshold and deadline of the task queue and has been standardized. Based on the priority of the resource requirements, the remaining resources are allocated proportionally to each service process to generate a resource allocation scheme that includes the memory allocation ratio, processor allocation ratio, and network bandwidth allocation ratio for different service processes. The service adaptation model simultaneously obtains task information for all currently pending transformation tasks. The task information includes the document type, document data volume, task submission time, and task recipient requirements. The processing difficulty is determined based on the document type in the task information, the processing time is determined based on the document data volume, and the urgency is determined based on the task submission time and recipient requirements. Based on processing difficulty, processing time and urgency, the service adaptation model establishes a task priority evaluation system, scores the priority of each conversion task to be processed, and determines the task priority sorting method according to the score from high to low, forming the priority sorting part in the task scheduling rules. The overall processing capability score of each service process is calculated based on its historical processing efficiency, real-time resource utilization, and the types of documents it supports. Set resource availability thresholds and processing capacity scoring thresholds; Tasks whose priority meets the preset sorting criteria are assigned to service processes whose comprehensive processing capability score is higher than the processing capability score threshold and whose real-time resource occupancy rate is lower than the resource availability rate threshold, forming the task allocation method part of the task scheduling rule, thus completing the generation of the resource allocation scheme and task scheduling rule.

6. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 1, characterized in that, The step of inputting the service stability control strategy into the execution control module of the document conversion service, executing the service stability control strategy, and obtaining the service operation parameter adjustment results includes: The service stability control strategy is input into the execution control module of the document conversion service through the data transmission channel. The execution control module has a built-in strategy parsing unit, which performs structured parsing on the service stability control strategy and extracts the resource allocation scheme and task scheduling rules. Based on the parsed resource allocation scheme, the execution control module sends a resource adjustment instruction to the resource allocation module of the document conversion service. The resource adjustment instruction includes the target memory usage, target processor usage, and target network bandwidth usage of each service process. After receiving the instruction, the resource allocation module gradually adjusts the resource usage of each service process so that the final resource allocation ratio conforms to the resource allocation scheme. Based on the parsed task scheduling rules, the execution control module sends a task adjustment instruction to the task scheduling module of the document conversion service. The task adjustment instruction contains a task priority sorting list and the target service process identifier corresponding to each task. After receiving the task adjustment instruction, the task scheduling module reorders the current queue of tasks to be processed and assigns the tasks to the corresponding service processes according to the sorting result. During the process of resource adjustment by the resource allocation module and task adjustment by the task scheduling module, the execution control module collects service operation parameter change data through the real-time monitoring unit. The service operation parameter change data includes resource usage change data, task processing efficiency change data, and process response change data. The collection frequency is adapted to the change rate of the service operation parameters. The execution control module performs time-series processing on the collected service operation parameter change data, sorts the service operation parameter change data according to the collection time order, and labels the attributes of the valid data, including the corresponding service process, data type and collection time. After processing, the service operation parameter adjustment result is generated, which reflects the change in service operation status after the service stability control strategy is implemented.

7. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 6, characterized in that, During the process of resource adjustment by the resource allocation module and task adjustment by the task scheduling module, the execution control module collects change data of service operation parameters through the real-time monitoring unit. This change data includes changes in resource usage, task processing efficiency, and process response. The collection frequency is adapted to the rate of change of the service operation parameters, including: The real-time monitoring unit of the execution control module determines the sampling frequency. The sampling frequency is determined based on the rate of change of historical service operation parameters. Parameters with a fast rate of change use a more frequent sampling frequency, while parameters with a slow rate of change use a relatively gentle sampling frequency, so that the sampling frequency is adapted to the rate of change of the parameters. During the resource allocation module's resource adjustment process, the real-time monitoring unit collects resource usage data for each service process at a determined collection frequency. After each collection, it calculates the difference between the current resource usage data and the previously collected resource usage data for that service process, associates the difference with the collection timestamp, and standardizes the change values ​​of different types of resources to form resource usage change data. The resource usage change data includes standardized memory usage change values, processor usage change values, and network bandwidth usage change values. During the task scheduling module's task adjustment process, the real-time monitoring unit collects the number of tasks completed by each service process within a unit of time, calculates the difference between the current unit of time and the previous unit of time, and obtains task processing efficiency change data. At the same time, the real-time monitoring unit collects the response time from receiving a request to returning a processing result for each service process, calculates the difference between the current response time and the previous response time, and obtains process response change data. The real-time monitoring unit verifies the collected data on changes in resource usage, task processing efficiency, and process response in real time. The verified data are then stored according to the service process, with each service process corresponding to a complete set of change data.

8. The method for optimizing the stability of document conversion services based on artificial intelligence according to claim 1, characterized in that, The step of updating the training sample set of the service adaptation model based on the service operation parameter adjustment results, and iteratively optimizing the service adaptation model to generate an optimized service adaptation model includes: The service operation parameter adjustment results are analyzed, and the resource usage change data, task processing efficiency change data, and process response change data are extracted. At the same time, the service anomaly pattern identification results and service stability control strategies corresponding to this stability control process are retrieved. The resource usage change data, task processing efficiency change data, process response change data, service anomaly pattern identification results, and service stability control strategies are associated with each other, based on the timestamp of the control process. The associated dataset is labeled as a new training sample. The new training sample contains input features and output labels. The input features are the abnormal association features in the service anomaly pattern recognition results. The output labels are the service stability control strategy and the corresponding service operation parameter adjustment results. The new training sample is added to the training sample set of the service adaptation model, and the contents of the training sample set are updated. From the updated training sample set, samples are selected as verification samples according to a preset ratio, and the remaining samples are used as training samples. The preset ratio is determined according to the total number of training samples, so that the verification samples can cover samples corresponding to different types of service anomalies. The training samples are divided into batches, each batch containing a preset number of training samples. After the training samples are input into the service adaptation model in the order of the batches, the service adaptation model processes the input features in each training sample and outputs a predicted service stability control strategy. The predictive control strategy output by the service adaptation model is compared with the actual service stability control strategy in the training samples. The differences between the two in terms of resource allocation ratio and task scheduling rules are calculated to generate a strategy difference value. At the same time, the parameter change results simulated by the service adaptation model based on the predictive control strategy are compared with the actual service operation parameter adjustment results in the training samples. Before the comparison, the parameter change results are standardized, and the differences between the two in terms of standardized resource consumption change, task processing efficiency change, and process response change are calculated to generate a parameter difference value. A loss function is constructed based on the strategy difference value and the parameter difference value. The partial derivatives of the loss function with respect to each parameter of the service adaptation model are calculated using the gradient descent algorithm. The parameters of the service adaptation model are adjusted according to the direction of the partial derivatives. The loss function value is adjusted to a preset range to complete the training of one batch of samples. This process is repeated until all training sample batches are processed to complete one round of training. After each round of training, the validation samples are input into the currently trained service adaptation model. The service adaptation model outputs the prediction and control strategy and the prediction parameter change results for the input features of the validation samples. The strategy difference between the prediction and control strategy and the actual service stability control strategy in the validation samples is calculated, and the parameter difference between the prediction parameter change results and the actual service operation parameter adjustment results in the validation samples is calculated. The validation accuracy of the service adaptation model is calculated based on the policy difference value and parameter difference value of the validation samples. The validation accuracy reflects the degree of agreement between the prediction results of the service adaptation model and the actual results. If the validation accuracy does not reach the preset standard, the training parameters are adjusted and the training samples are retrained according to the batch processing until the validation accuracy of the service adaptation model reaches the preset standard, thus completing the training optimization of the service adaptation model.

9. A document conversion service stability optimization system based on artificial intelligence, characterized in that, The AI-based document conversion service stability optimization system includes a processor and a memory, the memory and the processor being connected. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the AI-based document conversion service stability optimization method according to any one of claims 1-8.

Images