Intelligent instruction arrangement method for domestic operating system

Abstract

The invention relates to the technical field of computers, and discloses an intelligent instruction arrangement method for a domestic operating system, which comprises the following steps of: collecting operation instructions and related data of the domestic operating system through interface calling, log files and network transmission modes, and generating original operation instruction data, the related data comprises operating system types, instruction contents and execution result information; performing cleaning, duplicate removal and classification preprocessing operation on the original operation instruction data to generate preprocessed operation instruction data; by means of interface calling, log file and network transmission modes, original operation instructions and related context information of different domestic operation systems can be comprehensively obtained in real time, the integrity and real-time performance of multi-source instruction data are guaranteed, collected data are stored in a unified mode, and a summarizing report is generated, so that the summarizing efficiency is improved. A complete and reliable data basis is provided for subsequent processing, and the comprehensiveness and reliability of input arrangement are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a kind of instruction intelligent arrangement method of domestic operating system. BACKGROUND

[0002] Domestic operating system is a self-controllable operating system based on Linux kernel secondary development, with the rapid development and popularization of information technology and Internet, e-commerce has become an irresistible modern business trend, domestic operating system is gradually widely used in various industries in computer field.

[0003] At present, due to the existence of various system differences and complexity in the instruction arrangement process of domestic operating system, when the intelligent arrangement of cross-system operation instruction is carried out, it is impossible to collect multi-source instruction data in real time, and there is lack of efficient cleaning, deduplication and classification mechanism in data preprocessing process, so that the original instruction data has noise, repetition and inconsistent format problem, which will cause the analysis foundation to be not firm, and affect the accuracy of arrangement.

[0004] Therefore, the present application provides an instruction intelligent arrangement method of domestic operating system to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides an instruction intelligent arrangement method of domestic operating system, which solves the problems raised in the background art.

[0006] To achieve the above purpose, the present application provides the following technical scheme: an instruction intelligent arrangement method of domestic operating system, the method comprises the following steps:

[0007] S1, collect the operation instruction and related data of domestic operating system through interface call, log file, network transmission mode, and generate original operation instruction data, the related data includes operating system type, instruction content, execution result information;

[0008] S2, clean, deduplicate and classify the original operation instruction data to generate preprocessed operation instruction data;

[0009] S3, depth analysis is carried out on the preprocessed operation instruction, including analysis of instruction content, execution result and system response time, the syntax structure of instruction is analyzed through syntax parser, the semantic features of instruction are extracted through semantic analyzer, the operation instruction difference and characteristics between different systems are identified, and analyzed operation instruction data is generated;

[0010] S4, training and learning the analyzed operation instruction data by using a machine learning algorithm and an artificial intelligence technology, and constructing an intelligent arrangement model of cross-system operation instructions, the model being trained by using a neural network, a decision tree, and a support vector machine algorithm;

[0011] S5, intelligently arranging new operation instructions according to the intelligent arrangement model, automatically generating operation instructions suitable for target systems based on a matching result of an instruction feature vector, and adjusting an execution order of the instructions according to a topological sorting result to generate arranged operation instructions;

[0012] S6, performing actual execution and effect evaluation on the arranged operation instructions, comparing a numerical difference between an actual execution result and an expected result by using a difference calculation function, and adjusting and optimizing parameters of the arrangement model according to a gradient descent algorithm.

[0013] Preferably, the operation instructions and related data of the domestic operating system collected in S1 include the following steps:

[0014] S11, collecting operation instruction data by using an application programming interface of the domestic operating system, the operation instruction data including system call instructions and user commands;

[0015] S12, collecting execution result information by analyzing log files of the domestic operating system, the execution result information including instruction execution success and failure states and time stamps;

[0016] S13, capturing instruction flow data between the domestic operating systems in real time by using a network transmission protocol, the instruction flow data including source system types, target system types, and transmission protocol types;

[0017] S14, storing the collected operation instructions and related data in a distributed database, and generating a data collection summary report containing data amount statistical information.

[0018] Preferably, the preprocessing operation on the original operation instruction data in S2 includes the following steps:

[0019] S21, performing cleaning processing on the collected operation instruction data, removing duplicate instructions, invalid data, and noise data, and performing data validity verification based on a preset rule engine;

[0020] S22, performing a deduplication operation on the cleaned data, calculating a hash value of instruction content by using a hash algorithm, and performing uniqueness judgment and duplicate instruction elimination based on a hash value comparison result;

[0021] S23, performing classification processing on the deduplicated data, classifying the data according to instruction types, operating system types, and execution priorities, and generating corresponding classification labels;

[0022] S24, converting the classified data into a structured format and storing in a data warehouse of a columnar storage structure.

[0023] Preferably, the deep analysis of the pre-processed operation instruction in S3 comprises the following steps:

[0024] S31, instruction content analysis, using a syntax parser to parse the syntax structure of the instruction, using natural language processing technology to identify instruction keywords, parameters and operators;

[0025] S32, execution result analysis, statistics of instruction execution success rate, failure rate, error code distribution, generating an execution result report containing statistical indicators;

[0026] S33, system response time analysis, measuring the time delay of the instruction from sending to response, calculating the average response time, peak delay and delay variance;

[0027] S34, based on syntax and semantic analysis, extracting the semantic features of the instruction, constructing an instruction feature vector, the feature vector including instruction complexity, dependency relationship, resource occupancy rate;

[0028] S35, grouping the instruction feature vector by K-means clustering algorithm, identifying the instruction mode difference between different systems based on the distance of cluster center.

[0029] Preferably, the S4 of constructing the intelligent orchestration model of cross-system operation instruction comprises the following steps:

[0030] S41, preparing a training data set, dividing the analyzed operation instruction data into training set, validation set and test set, using k-fold cross-validation method to determine the division ratio;

[0031] S42, selecting machine learning algorithms, including neural network, decision tree, support vector machine, initializing the model, setting learning rate, iteration number of hyperparameters;

[0032] S43, using the training set to train the model, for neural network training, using back propagation algorithm to update the weight, the weight update formula is:

[0033] ;

[0034] wherein, is the weight change, represents the weight change between the th input and the th neuron in the neural network, represents the learning rate, represents the error term, represents the th input feature value, as an input feature index, as a neuron index;

[0035] S44, adjust the hyperparameters using the validation set, prevent overfitting using the early stopping method, calculate the model accuracy, recall rate and F1 score using the test set;

[0036] S45, save the trained model and generate a model performance report containing performance indicator values.

[0037] Preferably, the intelligent arrangement of new operation instructions in S5 includes the following steps:

[0038] S51, input the new operation instruction to the trained intelligent arrangement model, the new instruction includes real-time stream instruction and batch instruction;

[0039] S52, the model performs inference according to the instruction feature vector, outputs the arranged instruction sequence, and the sequence generates a sequence-to-sequence model based on the encoder-decoder structure;

[0040] S53, optimize the instruction execution order, calculate the shortest path of the instruction dependency graph using the path search algorithm, and determine the instruction execution order based on the path weight;

[0041] S54, generate the final arrangement instruction set, convert the instruction format according to the syntax specification of the target domestic operating system, and output the arrangement result file that meets the requirements of the target system.

[0042] Preferably, the actual execution and effect evaluation of the arranged operation instructions in S6 include the following steps:

[0043] S61, execute the arranged operation instructions in the actual domestic operating system environment, and execute the instructions through the system simulator and the physical deployment environment;

[0044] S62, collect execution result data, including output data, error log, performance indicator;

[0045] S63, compare the actual execution result with the expected result, and calculate the difference index using the root mean square error formula;

[0046] S64, adjust the model parameters based on the difference, fine-tune the model using the stochastic gradient descent method, and retrain the output layer parameters of the model;

[0047] S65, evaluate the performance of the optimized model, and when the loss function value does not reach the preset threshold, repeat the parameter adjustment step until the model converges.

[0048] Preferably, the construction of the instruction feature vector in S34 includes the following steps:

[0049] S341. Extract the syntactic features of the instruction, including instruction length, number of operators, and parameter types;

[0050] S342. Extract the semantic features of the instructions and use the Word2Vec word embedding model to map the instructions into a 128-dimensional vector;

[0051] S343 combines grammatical and semantic features to generate a 256-dimensional feature vector, and uses the Min-Max normalization method for data standardization.

[0052] Preferably, the neural network training in S43 includes the following steps:

[0053] S431. A multilayer perceptron structure is adopted, including an input layer, a hidden layer, and an output layer. The forward propagation calculation network output is based on the input feature vector and the current weights to calculate the activation values ​​of each layer. The forward propagation formula is expressed as:

[0054] ;

[0055] ;

[0056] in, Indicates the first The weighted sum of neurons, Indicates the first The input and the first The weights between neurons Indicates the first Each input feature value Indicates the first Bias terms for each neuron, Indicates the first The activation output of each neuron Represents the ReLU activation function. For input feature index, For neuron indexing;

[0057] S432. Calculate the output layer error. Use the mean squared error function to compare the difference between the network output and the expected output. The mean squared error function is expressed as:

[0058] ;

[0059] in, This represents the value of the loss function. Indicates the first The actual network output for each sample Indicates the first The expected output for each sample Indicates the total number of samples;

[0060] S433, backpropagation error, calculate the error term of each neuron layer by layer from the output layer to the input layer The error term is calculated based on the gradient of the loss function.

[0061] S434, update weights and biases, according to the error term , learning rate And input value Adjust the weight And bias The weight update formula is:

[0062] ;

[0063] Where, The weight change represents the weight change between the first Input and the first Neuron in the neural network, Indicates the learning rate, Indicates the error term, Indicates the first Input feature value, Is the input feature index, Is the neuron index.

[0064] Preferably, the S53 instruction execution order optimization includes the following steps:

[0065] S531, construct an instruction dependency graph, nodes represent instructions, and edges represent dependencies;

[0066] S532, determine the instruction execution order using the Kahn topological sorting algorithm, and sort based on the node in-degree value;

[0067] S533, calculate the optimal execution path using the Dijkstra algorithm in the dynamic programming algorithm, with time consumption minimization as the objective function;

[0068] S534, verify the orchestration logic through a model detection tool, and output verification reports and error diagnosis information.

[0069] Compared with the prior art, the present application provides an intelligent instruction orchestration method for a domestic operating system, which has the following beneficial effects:

[0070] 1. In the present application, through interface call, log file, network transmission mode, the original operation instructions and related context information of different domestic operating systems can be comprehensively and real-time acquired, the integrity and real-time of multi-source instruction data are ensured, the collected data is uniformly stored and summary report is generated, which provides complete and reliable data basis for subsequent processing, ensures that the intelligent arrangement process can fully reflect the actual running characteristics of different systems, and improves the comprehensiveness and reliability of arrangement input.

[0071] 2. In the present application, through the preprocessing operation of cleaning, deduplication and classification of original operation instruction data, data validity verification is carried out based on preset rule engine, and hash algorithm is used for uniqueness judgment and repeated instruction elimination, so that the problems of invalid data, noise data and repeated instructions in original data are solved. The classified data is converted into structured format for storage, which provides high-quality data input for subsequent deep analysis and model training, reduces the negative influence of data quality problems on instruction analysis and arrangement accuracy, and ensures the accuracy and consistency of intelligent arrangement model learning basis.

[0072] 3. In the present application, by constructing an intelligent arrangement model integrating machine learning algorithm, and using instruction feature vector matching, topological sorting and sequence generation technology, the instruction differences between different domestic operating systems can be automatically identified and adapted, and the conversion and optimization sorting of new operation instructions to target system can be realized. Through executing the arrangement result in actual environment and carrying out effect evaluation, the model parameters are continuously optimized according to the difference degree feedback, forming a closed-loop improvement mechanism, improving the execution success rate and system compatibility of arrangement instructions, and ensuring the continuous optimization of arrangement effect. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 The flowchart of the intelligent arrangement method of the instruction of the domestic operating system of the present application. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0075] Specific embodiment: please refer to Figure 1 An intelligent arrangement method of instruction of a domestic operating system, the method comprising the following steps:

[0076] S1, collect the operation instructions and related data of the domestic operating system through interface calls, log files, and network transmission methods, and generate raw operation instruction data, the related data including operating system type, instruction content, and execution result information;

[0077] S2, perform cleaning, deduplication, and classification preprocessing operations on the raw operation instruction data to generate preprocessed operation instruction data;

[0078] S3, perform deep analysis on the preprocessed operation instructions, including analysis of instruction content, execution result, and system response time, analyze the syntax structure of the instructions through a syntax parser, extract the semantic features of the instructions through a semantic analyzer, identify the differences and features of operation instructions between different systems, and generate analyzed operation instruction data;

[0079] S4, use machine learning algorithms and artificial intelligence techniques to train and learn the analyzed operation instruction data, and build an intelligent scheduling model for cross-system operation instructions, the model trained using neural networks, decision trees, and support vector machine algorithms;

[0080] S5, according to the intelligent scheduling model, intelligently schedule new operation instructions, automatically generate operation instructions suitable for the target system based on the instruction feature vector matching results, and adjust the execution order of the instructions according to the topological sorting results to generate scheduled operation instructions;

[0081] S6, perform actual execution and effect evaluation on the scheduled operation instructions, compare the numerical differences between the actual execution results and the expected results through a difference calculation function, and adjust and optimize the parameters of the scheduling model according to the gradient descent algorithm.

[0082] The collection of operation instructions and related data of the domestic operating system in S1 includes the following steps:

[0083] S11, collect operation instruction data through the application programming interface of the domestic operating system, the operation instruction data including system call instructions and user commands;

[0084] S12, collect execution result information by analyzing the log files of the domestic operating system, including instruction execution success and failure status, and timestamp;

[0085] S13, capture instruction flow data between domestic operating systems in real time through network transmission protocols, including source system type, target system type, and transmission protocol type;

[0086] S14, store the collected operation instructions and related data in a distributed database, and generate a data collection summary report containing data volume statistical information.

[0087] The preprocessing operation on the original operation instruction data in S2 includes the following steps:

[0088] S21, cleaning the collected operation instruction data, removing repeated instructions, invalid data and noise data, and verifying the data validity based on a preset rule engine;

[0089] S22, de-duplication operation is performed on the cleaned data, hash value of instruction content is calculated by using hash algorithm, uniqueness judgment and repeated instruction elimination are performed based on hash value comparison result, and the specific steps include the following steps:

[0090] S221, the instruction content is converted into a fixed-length byte sequence, and the UTF-8 encoding format is used for encoding processing;

[0091] S222, the hash function is applied to iteratively calculate the byte sequence, each iteration round includes byte grouping, bit operation and modulo addition operation, and intermediate hash value is generated, and the hash function is represented as:

[0092] ;

[0093] Wherein, represents the hash value of the i-th iteration round, represents the hash value of the j-th iteration round, represents the i-th byte grouping, represents the iteration function, represents the iteration round index; S23, the de-duplicated data is classified, and the classification label is generated according to the instruction type, operating system type and execution priority;

[0094] S24, the classified data is converted into a structured format and stored in a columnar storage structure data warehouse.

[0095] S3, the preprocessed operation instruction is analyzed in depth, including the following steps:

[0096] S31, instruction content analysis is performed, syntax parser is used to parse the syntax structure of the instruction, natural language processing technology is used to identify instruction keywords, parameters and operators, and the specific steps include the following steps:

[0097] S311, the instruction content is segmented, and the instruction string is segmented into a word sequence by using a maximum matching algorithm based on a dictionary;

[0098] S311, the instruction content is segmented, and the instruction string is segmented into a word sequence by using a maximum matching algorithm based on a dictionary;

[0099] ​​S312, perform part-of-speech tagging, use a hidden Markov model to identify the part-of-speech category of each word, the state transition probability of the hidden Markov model is represented as:

[0100] ;

[0101] wherein, represents the part-of-speech state at time , represents the part-of-speech state at time , represents the probability of transitioning from state to , represents a frequency counting function, is a time index;

[0102] S313, parse the syntax structure of the instruction, identify the subject, predicate and object relationship through the dependency syntax analysis tree;

[0103] S314, extract semantic roles, use semantic role labeling technology to identify action, agent and patient elements in the instruction;

[0104] S32, perform execution result analysis, statistics the success rate, failure rate, error code distribution of instruction execution, generate an execution result report containing statistical indicators;

[0105] S33, perform system response time analysis, measure the time delay of the instruction from sending to response, calculate the average response time, peak delay and delay variance;

[0106] S34, based on syntax and semantic analysis, extract the semantic features of the instruction, construct an instruction feature vector, the feature vector includes instruction complexity, dependency relationship, resource occupancy rate;

[0107] S35, group the instruction feature vectors through the K-means clustering algorithm, identify the instruction mode differences between different systems based on the distance of the cluster centers, specifically including the following steps:

[0108] S351, initialize the cluster center, randomly select k vectors from the instruction feature vector set as the initial center point;

[0109] S352, calculate the Euclidean distance of each instruction feature vector to each cluster center, assign the vector to the nearest cluster, the Euclidean distance formula is represented as:

[0110] ;

[0111] wherein, represents the Euclidean distance between vector and cluster center , representing an instruction feature vector, representing a cluster center vector, representing a feature dimension;

[0112] S353, updating the cluster center, calculating the mean of all vectors in each cluster as a new center point;

[0113] S354, iteratively performing S352 and S353 steps until the cluster center changes less than a preset threshold and reaches a maximum number of iterations.

[0114] The intelligent arrangement model of the cross-system operation instruction in S4 includes the following steps:

[0115] S41, preparing a training data set, dividing the analyzed operation instruction data into a training set, a validation set and a test set, and determining the division ratio by using a k-fold cross-validation method;

[0116] S42, selecting a machine learning algorithm, including neural network, decision tree, support vector machine, initializing the model, and setting the learning rate, the number of iterations and the hyperparameters;

[0117] S43, training the model using the training set, for neural network training, using the back propagation algorithm to update the weight, and the weight update formula is:

[0118] ;

[0119] wherein, is the weight change, represents the weight change amount between the th input and the th neuron in the neural network, represents the learning rate, represents the error term, represents the th input feature value, is the input feature index, is the neuron index;

[0120] S44, adjusting the hyperparameters using the validation set, using the early stopping method to prevent overfitting, and using the test set to calculate the model accuracy, recall rate and F1 score;

[0121] S45, saving the trained model and generating a model performance report containing performance indicator values.

[0122] The intelligent arrangement of the new operation instruction in S5 includes the following steps:

[0123] S51, inputting the new operation instruction into the trained intelligent arrangement model, the new instruction including real-time stream instructions and batch instructions;

[0124] S52, the model performs inference according to the instruction feature vector, outputs the arranged instruction sequence, and the sequence generation is based on the sequence-to-sequence model of the encoder-decoder structure;

[0125] S53, the execution order of the instructions is optimized, the shortest path of the instruction dependency graph is calculated by using the path search algorithm, and the execution order of the instructions is determined based on the path weight;

[0126] S54, a final arranged instruction set is generated, the instruction format is converted according to the syntax specification of the target domestic operating system, and an arranged result file meeting the requirements of the target system is output.

[0127] S6, actual execution and effect evaluation of the arranged operation instructions include the following steps:

[0128] S61, the arranged operation instructions are executed in the actual domestic operating system environment, and the instructions are executed through the system simulator and the physical deployment environment;

[0129] S62, the execution result data including output data, error logs and performance indicators are collected;

[0130] S63, the actual execution result is compared with the expected result, and the difference index is calculated by using the root mean square error formula;

[0131] S64, the model parameters are adjusted based on the difference, the random gradient descent method is used for model fine-tuning, the output layer parameters of the model are retrained, and the specific steps include the following steps:

[0132] S661, a small batch of sample sets are randomly extracted from the training data;

[0133] S662, the loss function gradient of the small batch of samples is calculated, the gradient is calculated for each sample and the average value is taken, and the gradient update formula is represented as:

[0134] ;

[0135] wherein, represents the model parameters of the th iteration, represents the updated parameters, represents the learning rate, represents the loss function at the parameter gradient, is the number of iterations;

[0136] S663, the model parameters are updated, the weights are adjusted in the opposite direction of the gradient, and the step is controlled by the learning rate;

[0137] S664, steps S661 to S663 are repeated until the loss function converges and the maximum number of iterations is reached;

[0138] S65, evaluating the performance of the optimized model, when the loss function value does not reach the preset threshold, repeating the parameter adjustment step until the model converges.

[0139] The instruction feature vector constructed in S34 includes the following steps:

[0140] S341, extracting the syntax features of the instruction, including instruction length, operator number, and parameter type;

[0141] S342, extracting the semantic features of the instruction, using a Word2Vec word embedding model to map the instruction to a 128-dimensional vector, including the following steps:

[0142] S3421, constructing an instruction vocabulary, counting all words appearing in the instruction and assigning a unique index;

[0143] S3422, training word vectors, using a Skip-gram model to predict target words based on context words, the objective function of the Skip-gram model is represented as:

[0144] ;

[0145] wherein, represents the loss function, represents the model parameters, represents the length of the instruction sequence, represents the index of the target word, represents the index of the context word, represents the context window size, represents the conditional probability function, is the sequence position index, is the context offset index;

[0146] S3423, learning the distributed representation of words through a neural network, outputting a 128-dimensional real-valued vector for each word;

[0147] S3424, combining the word vectors corresponding to the word sequence in the instruction to generate the semantic feature vector of the instruction;

[0148] S343, combining the syntax and semantic features to generate a 256-dimensional feature vector, using the Min-Max normalization method for data standardization processing, the calculation formula is:

[0149] ;

[0150] wherein, represents any original feature value in the 256-dimensional feature vector, This indicates the minimum value of this feature dimension in the training dataset. This indicates the maximum value of this feature dimension in the training dataset. This represents the normalized eigenvalues.

[0151] The neural network training in S43 includes the following steps:

[0152] S431. A multilayer perceptron structure is adopted, including an input layer, a hidden layer, and an output layer. The forward propagation calculation network output is based on the input feature vector and the current weights to calculate the activation values ​​of each layer. The forward propagation formula is expressed as:

[0153] ;

[0154] ;

[0155] in, Indicates the first The weighted sum of neurons, Indicates the first The input and the first The weights between neurons Indicates the first Each input feature value Indicates the first Bias terms for each neuron, Indicates the first The activation output of each neuron Represents the ReLU activation function. For input feature index, For neuron indexing;

[0156] S432. Calculate the output layer error. Use the mean squared error function to compare the difference between the network output and the expected output. The mean squared error function is expressed as:

[0157] ;

[0158] in, This represents the value of the loss function. Indicates the first The actual network output for each sample Indicates the first The expected output for each sample Represents the total number of samples. Indicates the sample index;

[0159] S433, Backpropagation error: Calculate the error term for each neuron layer by layer from the output layer to the input layer. The error term is calculated based on the gradient of the loss function;

[0160] S434. Update weights and biases based on the error term. Learning rate and input values Adjust weights and bias The weight update formula is:

[0161] ;

[0162] in, The weight change represents the weight change in the neural network. The input and the first The amount of weight change between neurons Indicates the learning rate. Indicates the error term. Indicates the first Each input feature value For input feature index, For neuron indexing.

[0163] Optimizing the instruction execution order in S53 includes the following steps:

[0164] S531. Construct an instruction dependency graph, where nodes represent instructions and edges represent dependencies.

[0165] S532. Use the Kahn topological sorting algorithm to determine the order of instruction execution, sorting based on the in-degree value of nodes, specifically including the following steps:

[0166] S5321. Calculate the in-degree of each node and count the number of edges pointing to each node. The formula for calculating the in-degree is as follows:

[0167] ;

[0168] in, Represents a node in-degree, Represents a set of nodes. Indicates the source node, As an indicator function, when there exists an edge from arrive The value is 1 if it is true, and 0 otherwise.

[0169] S5322. Initialize the queue by adding all nodes with an in-degree of 0 to the queue;

[0170] S5323. Take a node from the queue, output the node, and decrease the in-degree of all its neighboring nodes.

[0171] S5324. When the in-degree of a neighboring node becomes 0, add it to the queue. Repeat step S783 until the queue is empty.

[0172] S533, calculate the optimal execution path by using the Dijkstra algorithm in the dynamic programming algorithm, with the time consumption minimization as the objective function, specifically comprising the following steps:

[0173] S5331, initialize the instruction dependency graph, take each instruction as a node, the dependency relationship as an edge, and assign a weight to each edge representing the execution time;

[0174] S5332, set the starting node as the current node, initialize the distance of all nodes to the starting node as infinity, and the distance of the starting node as 0;

[0175] S5333, traverse all neighbor nodes of the current node, update the shortest distance of the neighbor nodes to the starting node, and the distance update formula is represented as:

[0176]

[0177] wherein, represents the current shortest distance of the node to the starting node, represents the distance of the node to the starting node, represents the edge weight from the node to , and represents the minimum value function;

[0178] S5334, select the node with the smallest distance among the unvisited nodes as the new current node, and repeat step S5533 until all nodes are visited.

[0179] S534, verify the orchestration logic by using a model detection tool, and output a verification report and error diagnosis information.

[0180] The operation steps of the instruction intelligent orchestration method of the domestic operating system are as follows:

[0181] Step 1, multi-source data acquisition and preprocessing

[0182] Firstly, the operation instructions and related data of the domestic operating system are collected in real time through interface calling, log file parsing and network transmission, including operating system type, instruction content and execution result information, to generate raw operation instruction data. Subsequently, the raw operation instruction data is cleaned, de-duplicated and classified for preprocessing: the data validity is verified based on a rule engine, the hash value of the instruction content is calculated by using a hash algorithm to realize de-duplication, and the data is classified and labeled according to the instruction type, operating system type and execution priority. Finally, the classified data is converted into a structured format and stored in a columnar storage structure data warehouse to generate preprocessed operation instruction data, providing standardized input for subsequent analysis. ​

[0183] Step two, deep analysis and feature extraction

[0184] The pre-processed operation instruction data is subjected to multi-dimensional deep analysis. The syntax structure of the instruction is parsed by a syntax parser, and the instruction keywords, parameters and operators are identified in combination with natural language processing technology; the success rate, failure rate and error code distribution of instruction execution are counted to generate an execution result report; the system response time delay is measured, and the average response time and peak delay are calculated. Based on syntax and semantic analysis, the semantic features of the instruction are extracted, and an instruction feature vector containing instruction complexity, dependency relationship and resource occupancy rate is constructed. Further, the K-means clustering algorithm is used to group the instruction feature vectors, identify the differences in instruction patterns between different systems, and generate the analyzed operation instruction data.

[0185] Step three, intelligent scheduling model construction and training

[0186] Machine learning algorithms and artificial intelligence techniques are used to train and learn the analyzed operation instruction data, and an intelligent scheduling model for cross-system operation instructions is constructed. The k-fold cross-validation method is used to divide the training set, validation set and test set, and the hyperparameters of neural networks, decision trees and support vector machines are initialized. For neural network training, a multi-layer perceptron structure is used, the network output is calculated by forward propagation, the difference between actual output and expected result is compared using the mean square error function, and the weights and biases are updated based on the back propagation algorithm. Finally, the trained model is saved, and a performance report containing accuracy and recall rate indicators is generated.

[0187] Step four, instruction intelligent scheduling and optimization

[0188] According to the trained intelligent scheduling model, new operation instructions are intelligently scheduled. Real-time streaming instructions and batch instructions are input to the model, and based on the instruction feature vector matching results, a sequence-to-sequence model with an encoder-decoder structure is used to output the scheduled instruction sequence. When optimizing the instruction execution order, an instruction dependency graph is constructed, the Kahn topological sorting algorithm is used to determine the execution order based on the node in-degree value, and the Dijkstra algorithm in the dynamic programming algorithm is used to calculate the optimal execution path with the goal of minimizing time consumption. Finally, a set of scheduled instructions that meet the syntax specifications of the target system is generated.

[0189] Step five, execution effect evaluation and parameter optimization

[0190] The actual execution and effect evaluation are performed on the arranged operation instructions. The instructions are executed through the system simulator and the physical deployment environment, and output data, error logs and performance index execution result data are collected. The difference degree index of the actual execution result and the expected result is calculated by using the root mean square error formula, and the model parameters are adjusted based on the difference degree: the model is fine-tuned by using the stochastic gradient descent method, the output layer parameters are retrained, and iteration is performed until the loss function converges. Finally, the model parameters are continuously optimized, and the precision and system adaptability of the instruction arrangement are improved.

[0191] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one from another entity or action, without necessarily requiring or implying any actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0192] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, alternatives and variations can be made to these embodiments without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent instruction arrangement in a domestically developed operating system, characterized in that: The method includes the following steps: S1. Collect operation instructions and related data of the domestic operating system through interface calls, log files, and network transmission, and generate raw operation instruction data. The related data includes operating system type, instruction content, and execution result information. S2. Perform cleaning, deduplication, and classification preprocessing operations on the original operation instruction data to generate preprocessed operation instruction data; S3. Perform in-depth analysis on the preprocessed operation instructions, including analysis of instruction content, execution results, and system response time. Analyze the syntactic structure of the instructions through a syntax parser, extract the semantic features of the instructions through a semantic analyzer, identify the differences and features of operation instructions between different systems, and generate analyzed operation instruction data. S4. Using machine learning algorithms and artificial intelligence technology, the analyzed operation instruction data is trained and learned to construct an intelligent orchestration model for cross-system operation instructions. The model is trained using neural network, decision tree, and support vector machine algorithms. S5. Based on the intelligent orchestration model, new operation instructions are intelligently orchestrated. Based on the instruction feature vector matching results, operation instructions suitable for the target system are automatically generated, and the execution order of the instructions is adjusted according to the topology sorting results to generate orchestrated operation instructions. S6. Perform actual execution and effect evaluation on the orchestrated operation instructions, compare the numerical difference between the actual execution result and the expected result using the difference calculation function, and adjust and optimize the parameters of the orchestration model according to the gradient descent algorithm.

2. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The step S1, which involves collecting operating instructions and related data from the domestic operating system, includes the following steps: S11. Collect operation instruction data by calling the application programming interface of the domestic operating system. The operation instruction data includes system call instructions and user commands. S12. Collect execution result information by parsing the log files of the domestic operating system, including the success and failure status of instruction execution and timestamps; S13. Capture instruction stream data between domestic operating systems in real time through network transmission protocols, including source system type, target system type, and transmission protocol type; S14. Store the collected operation instructions and related data in a distributed database, and generate a data collection summary report containing data volume statistics.

3. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The preprocessing operation on the original operation instruction data in step S2 includes the following steps: S21. Clean the collected operation instruction data to remove duplicate instructions, invalid data and noisy data, and verify the validity of the data based on the preset rule engine. S22. Perform deduplication on the cleaned data, calculate the hash value of the instruction content using a hash algorithm, and perform uniqueness judgment and duplicate instruction removal based on the hash value comparison results. S23. Classify the deduplicated data according to instruction type, operating system type, and execution priority, and generate corresponding classification labels; S24. Convert the categorized data into a structured format and store it in a columnar storage data warehouse.

4. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The in-depth analysis of the preprocessed operation instructions in step S3 includes the following steps: S31. Perform instruction content analysis, use a syntax parser to parse the syntax structure of the instruction, and use natural language processing technology to identify instruction keywords, parameters and operators; S32. Analyze the execution results, calculate the success rate, failure rate, and error code distribution of the executed instructions, and generate an execution result report containing statistical indicators. S33. Perform system response time analysis, measure the time delay from command to response, and calculate the average response time, peak delay, and delay variance. S34. Based on syntax and semantic analysis, extract the semantic features of the instructions and construct the instruction feature vector, which includes instruction complexity, dependency relationship and resource utilization. S35. The instruction feature vectors are grouped using the K-means clustering algorithm, and the differences in instruction patterns between different systems are identified based on the distance between the cluster centers.

5. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The intelligent orchestration model for cross-system operation instructions in S4 includes the following steps: S41. Prepare the training dataset. Divide the analyzed operation instruction data into a training set, a validation set, and a test set. Use the k-fold cross-validation method to determine the division ratio. S42. Select a machine learning algorithm, including neural networks, decision trees, and support vector machines, initialize the model, and set hyperparameters such as learning rate and number of iterations. S43. Train the model using the training set. For neural network training, use the backpropagation algorithm to update the weights. The weight update formula is: ； in, The weight change represents the weight change in the neural network. The input and the first The amount of weight change between neurons Indicates the learning rate. Indicates the error term. Indicates the first Each input feature value For input feature index, For neuron indexing; S44. Use the validation set to adjust hyperparameters, use early stopping to prevent overfitting, and use the test set to calculate the model's accuracy, recall, and F1 score. S45. Save the trained model and generate a model performance report containing performance metric values.

6. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The intelligent arrangement of new operation instructions in S5 includes the following steps: S51. Input new operation instructions into the trained intelligent orchestration model, the new instructions including real-time stream instructions and batch instructions; S52. The model performs inference based on the instruction feature vector and outputs an arranged instruction sequence, which generates a sequence-to-sequence model based on an encoder-decoder structure. S53. Optimize the instruction execution order by using a path search algorithm to calculate the shortest path in the instruction dependency graph and determining the instruction execution order based on the path weight. S54. Generate the final orchestration instruction set, convert the instruction format according to the syntax specification of the target domestic operating system, and output the orchestration result file that meets the requirements of the target system.

7. The intelligent instruction orchestration method for a domestically developed operating system according to claim 1, characterized in that: The actual execution and effect evaluation of the arranged operation instructions in step S6 includes the following steps: S61. Execute the orchestrated operation instructions in a real domestic operating system environment, and execute the instructions through a system simulator and physical deployment environment; S62. Collect execution result data, including output data, error logs, and performance metrics; S63. Compare the actual results with the expected results, and calculate the difference index using the root mean square error formula. S64. Adjust the model parameters based on the difference, use stochastic gradient descent to fine-tune the model, and retrain the output layer parameters of the model. S65. Evaluate the performance of the optimized model. If the loss function value does not reach the preset threshold, repeat the parameter adjustment steps until the model converges.

8. The intelligent instruction orchestration method for a domestically developed operating system according to claim 4, characterized in that: The construction of the instruction feature vector in S34 includes the following steps: S341. Extract the syntactic features of the instruction, including instruction length, number of operators, and parameter types; S342. Extract the semantic features of the instructions and use the Word2Vec word embedding model to map the instructions into a 128-dimensional vector; S343 combines grammatical and semantic features to generate a 256-dimensional feature vector, and uses the Min-Max normalization method for data standardization.

9. The intelligent instruction orchestration method for a domestically developed operating system according to claim 5, characterized in that: The neural network training in S43 includes the following steps: S431. A multilayer perceptron structure is adopted, including an input layer, a hidden layer, and an output layer. The forward propagation calculation network output is based on the input feature vector and the current weights to calculate the activation values ​​of each layer. The forward propagation formula is expressed as: ； ； in, Indicates the first The weighted sum of neurons, Indicates the first The input and the first The weights between neurons Indicates the first Each input feature value Indicates the first Bias terms for each neuron, Indicates the first The activation output of each neuron Represents the ReLU activation function. For input feature index, For neuron indexing; S432. Calculate the output layer error. Use the mean squared error function to compare the difference between the network output and the expected output. The mean squared error function is expressed as: ； in, This represents the value of the loss function. Indicates the first The actual network output for each sample Indicates the first The expected output for each sample Indicates the total number of samples; S433, Backpropagation error: Calculate the error term for each neuron layer by layer from the output layer to the input layer. The error term is calculated based on the gradient of the loss function; S434. Update weights and biases based on the error term. Learning rate and input values Adjust weights and bias The weight update formula is: ； in, The weight change represents the weight change in the neural network. The input and the first The amount of weight change between neurons Indicates the learning rate. Indicates the error term. Indicates the first Each input feature value For input feature index, For neuron indexing.

10. The intelligent instruction orchestration method for a domestically developed operating system according to claim 6, characterized in that: The optimized instruction execution order in S53 includes the following steps: S531. Construct an instruction dependency graph, where nodes represent instructions and edges represent dependencies. S532. Use the Kahn topological sorting algorithm to determine the order of instruction execution, and sort based on the in-degree value of the nodes; S533. Use Dijkstra's algorithm in dynamic programming to calculate the optimal execution path, with the goal of minimizing time consumption. S534. Verify the orchestration logic using model testing tools, and output a verification report and error diagnosis information.

Images