Web service comprehensive score calculation method based on OWL-S

Abstract

The invention discloses an OWL-S-based Web service comprehensive score calculation method. The method comprises the following steps: using OWL-S semantics to describe Web services; selecting a multi-dimensional evaluation index; an evaluation model based on OWL-S is established; calculating a comprehensive score; optimizing the evaluation model by using machine learning; constructing a dynamic evaluation model; fusing domain knowledge; and a privacy protection mechanism is adopted. By combining the OWL-S semantic description language, multiple evaluation indexes, a machine learning algorithm, a dynamic evaluation mechanism and personalized recommendation, comprehensive evaluation and ranking of Web services are realized, the intelligent level of Web service evaluation is further improved, the evaluation accuracy, efficiency and user experience can be effectively improved, and the user experience is improved. And the requirements of Web service providers, users and platforms are met.

Description

Technical Field

[0001] The present invention belongs to the technical field of Web services, and in particular relates to a comprehensive scoring calculation method for Web services based on OWL-S. Background Art

[0002] With the widespread application of Web services, Web service evaluation has become increasingly important. Existing Web service evaluation methods are mainly based on Quality of Service (QoS) indicators such as response time, availability, and security, but these methods have some shortcomings, including the following aspects:

[0003] 1. Single evaluation index: Existing methods mainly focus on quality of service (QoS) indicators such as response time, throughput, availability and security, but lack consideration of service functions, user experience and other aspects, resulting in one-sided evaluation results that cannot fully reflect the performance and quality of Web services.

[0004] 2. Evaluation results lack reliability and consistency: There are large differences in the results obtained by different evaluation methods and evaluation subjects. There is a lack of unified evaluation standards and specifications, which makes it difficult to provide a reliable basis for service selection and optimization and cannot meet actual application needs.

[0005] 3. Lack of flexibility and adaptability: Existing methods generally use static evaluation indicators and models, which cannot automatically adjust the evaluation strategy according to real-time data such as user feedback and service quality changes, and are difficult to adapt to the dynamic changes and personalized needs of Web services.

[0006] 4. Lack of adaptability to emerging service models and application scenarios: With the in-depth application of Web services in various industries, Web services in specific fields such as medical care and finance have emerged. The existing general evaluation methods are difficult to meet the professional needs of these fields and lack effective evaluation means.

[0007] 5. Lack of personalized recommendation function: Existing methods cannot provide personalized Web service recommendations based on user preferences and needs. Users need to compare and choose on their own, which increases the user's decision-making cost and the user's experience in selecting Web services is poor.

[0008] 6. Insufficient privacy protection: Existing methods do not take effective privacy protection measures during the evaluation process. Users’ private information may be leaked or abused, affecting user trust. The evaluation process lacks a privacy protection mechanism, and user privacy security is threatened.

[0009] It can be seen that the existing Web service evaluation methods have problems such as single evaluation indicators, lack of reliability of results, lack of flexibility and personalization, inability to meet emerging needs, and insufficient privacy protection. They are unable to meet the growing needs of Web service evaluation. Summary of the invention

[0010] In order to address the deficiencies in the prior art, the present invention aims to provide a comprehensive scoring calculation method for Web services based on OWL-S. The method can achieve all-round evaluation and ranking of Web services by combining OWL-S semantic description language, multiple evaluation indicators, machine learning algorithms, dynamic evaluation mechanisms and personalized recommendations, thereby effectively solving the problems of single evaluation indicators, lack of reliability, flexibility and personalization in the existing Web service evaluation methods, and providing more accurate, reliable and efficient evaluation results.

[0011] In order to achieve the above object, the technical solution adopted by the present invention is:

[0012] A method for calculating a comprehensive score of a Web service based on OWL-S is implemented based on the following steps, including:

[0013] 1) OWL-S semantic description: Use OWL-S semantic description language to describe Web services;

[0014] 2) Multi-dimensional evaluation indicators: Select multiple evaluation indicators to conduct a comprehensive evaluation of Web services;

[0015] 3) Evaluation model establishment: According to OWL-S semantic description and multi-dimensional evaluation indicators, an evaluation model based on OWL-S is established;

[0016] 4) Evaluation result calculation: Use the evaluation model to evaluate and rank Web services and calculate a comprehensive score to reflect the overall performance and quality of the Web services;

[0017] 5) Machine learning optimization: Use machine learning algorithms to optimize and adjust the evaluation model and build a dynamic evaluation model;

[0018] 6) Dynamic evaluation model based on machine learning: Based on user feedback and real-time data on service quality changes, the dynamic evaluation model is used to automatically adjust the evaluation index weights and dynamically evaluate and rank Web services;

[0019] 7) Cross-domain knowledge integration: Integrate Web service evaluation with professional domain knowledge to provide professional and accurate evaluation for Web services in specific professional fields;

[0020] 8) Privacy protection mechanism: A privacy protection mechanism is adopted during the evaluation process to ensure user privacy security.

[0021] Furthermore, the OWL-S semantic description in step 1) includes service overview, service model, service foundation, service function, service interface and service quality;

[0022] The service overview is used to describe the basic information of the service, including the service name, provider, and function description;

[0023] The service model is used to describe the internal workflow and composition of the service;

[0024] The service foundation is used to describe how to access the service, including communication protocols and message formats.

[0025] Furthermore, the multidimensional evaluation indicators in step 2) include functional indicators, performance indicators, reliability indicators, safety indicators, and user experience indicators;

[0026] The functional indicators include functional integrity, accuracy, and applicability;

[0027] The performance indicators include response time, throughput, and resource utilization;

[0028] The reliability indicators include availability, fault tolerance, and recovery capability;

[0029] The security indicators include confidentiality, integrity, and authentication;

[0030] The user experience indicators include ease of use, learnability, and satisfaction.

[0031] Furthermore, in step 3), the evaluation model is constructed using AHP or fuzzy comprehensive evaluation method.

[0032] When the AHP analytic hierarchy process is used to construct an evaluation model, the following steps are included:

[0033] 3-1) Establish a hierarchical structure: including the goal layer, the criteria layer and the indicator layer; the goal layer is used to calculate the comprehensive score, the criteria layer is used to evaluate functionality, performance, and reliability, and the indicator layer is used to set specific indicators, including functional integrity, response time, and availability;

[0034] 3-2) Construct a judgment matrix: compare the elements of each level in pairs to determine their relative importance; and assuming that there are n evaluation indicators, construct an n×n judgment matrix A, where element a ij Indicates the importance of indicator i relative to indicator j;

[0035] 3-3) Calculate weights: Use the eigenvector method to calculate the weights of each indicator, including:

[0036] 3-3-1) Calculate the maximum eigenvalue λ of the judgment matrix max and its corresponding eigenvector α;

[0037] 3-3-2) Normalize the feature vector α to get the weight vector of each indicator:

[0038] w=(w1,w2,...,w n ),

[0039] in, γ=1,…,n;w1,w2,…,w n Respectively represent the weight of each indicator in the evaluation model; n represents the total number of evaluation indicators;

[0040] 3-4) Consistency check: ensure the consistency of judgment;

[0041] Calculate the consistency index CI and consistency rate CR:

[0042]

[0043]

[0044] Among them, RI is the random consistency index; when CR<0.1, the judgment matrix has satisfactory consistency.

[0045] Furthermore, the machine learning algorithm in step 5) includes SVM support vector machine or random forest algorithm.

[0046] The optimization of the evaluation model using SVM support vector machine specifically includes:

[0047] 5-1) Data preparation: Collect historical evaluation data, including the values ​​of each indicator and the final score;

[0048] 5-2) Feature selection: Use principal component analysis (PCA) to select the most important features;

[0049] 5-3) Model training: Use the training data set to train the SVM model; assuming that the training data set is where x i is the feature vector, y i is the score, m represents the number of training samples;

[0050] The SVM model is trained by optimizing the following objective function:

[0051]

[0052] Subject to y i (w T x i +b)≥1-ξ i ,ξ i ≥0,i=1,...,m

[0053] Among them, w is the weight vector, w T represents the transposed matrix of w, b is the bias, ξ iis the slack variable, C is the penalty parameter;

[0054] 5-4) Model validation: Use the validation dataset to evaluate model performance;

[0055] 5-5) Model application: Use the trained model to evaluate new web services and output a comprehensive score.

[0056] Furthermore, in step 6), a Q-learning reinforcement learning algorithm is used to construct a dynamic evaluation model, including:

[0057] 6-1) State definition: define the indicator values ​​of the current Web service as the state vector s;

[0058] 6-2) Action definition: Adjust the weight of each indicator as action a;

[0059] 6-3) Reward function: Define reward r based on user feedback and service quality changes;

[0060] 6-4) Q value update: Continuously update the Q value based on state-action-reward:

[0061]

[0062] Among them, α is the learning rate, which controls the influence of new information on the Q value update; γ is the discount factor, which is used to balance the immediate reward and long-term benefits; s′ is the new state after the action is performed; a′ represents the learning rate after the action is performed; Q(s,a) is the Q value of taking action a in state s, which represents the estimation of long-term benefits; represents the maximum Q value of all possible actions in the new state s';

[0063] 6-5) Strategy selection: Use the ε-strategy greedy strategy to select the optimal action;

[0064] The ε-greedy strategy includes: randomly selecting an action with probability ε, that is, exploration; selecting the action with the largest current Q value with probability 1-ε, that is, utilization; the formula is expressed as:

[0065] The probability P(a) of choosing action a is:

[0066]

[0067] Among them, |A| is the size of the action space, and Q(s,a′) is the Q value of action a′ in state s.

[0068] Furthermore, the privacy protection mechanism in step 8) adopts federated learning technology, which specifically includes:

[0069] 8-1) Data decentralization: Evaluation data is decentralized and stored among various participants;

[0070] 8-2) Local training: Each participant uses local data to train a local model;

[0071] 8-3) Model aggregation: The central server aggregates the local model parameters;

[0072] 8-4) Model update: Distribute the aggregated model to all participants.

[0073] The beneficial effects of the present invention are:

[0074] 1. The present invention aims to provide a comprehensive, accurate and reliable Web service evaluation method. By comprehensively considering multiple dimensions such as service functions, service quality, and user experience, a comprehensive evaluation index system is constructed to provide objective and fair evaluation results, thereby solving the problem that the existing methods have single evaluation indicators and one-sided results.

[0075] 2. The present invention improves the efficiency and automation level of Web service evaluation. By using technologies such as OWL-S semantic description and machine learning algorithms, the automatic evaluation and ranking of Web services can be realized, the cost and time of manual evaluation can be reduced, and the evaluation efficiency can be improved.

[0076] 3. The present invention enhances the adaptability and flexibility of Web service evaluation. It introduces a dynamic evaluation mechanism, automatically adjusts the evaluation index weights based on real-time data such as user feedback and service quality changes, and adapts to the dynamic characteristics of Web services. At the same time, it provides personalized evaluation solutions for different fields and different users.

[0077] 4. The present invention aims to comprehensively improve and optimize Web service evaluation in terms of comprehensiveness, accuracy, efficiency, adaptability, applicability and standardization of evaluation, provide support and guarantee for the healthy development of Web services, and ultimately serve stakeholders such as Web service providers, users and platforms.

[0078] 5. The present invention enriches the application scenarios of Web service evaluation. It provides diversified evaluation services and products for different entities such as Web service providers, service users, and service platforms to meet the evaluation needs in different scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 is a flow chart of a method for calculating a comprehensive score of a Web service according to the present invention; DETAILED DESCRIPTION

[0080] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not intended to limit the scope of use of the present invention.

[0081] like Figure 1As shown, the present invention proposes a method for calculating the comprehensive score of Web services based on OWL-S, and the specific steps include:

[0082] 1) OWL-S semantic description: Use OWL-S semantic description language to describe Web services. OWL-S semantic description includes service overview, service model, service foundation, service function, service interface and service quality.

[0083] Service overview, which is used to describe the basic information of the service, including the service name, provider, function description, etc.;

[0084] Service model, which is used to describe the internal workflow and composition of the service;

[0085] Service foundation, which is used to describe how to access the service, including communication protocols, message formats, etc.

[0086] 2) Multi-dimensional evaluation indicators: Select multiple evaluation indicators to conduct a comprehensive evaluation of Web services, including functional indicators, performance indicators, reliability indicators, security indicators, user experience indicators, etc.

[0087] Functional indicators include functional completeness, accuracy, applicability, etc.; performance indicators include response time, throughput, resource utilization, etc.; reliability indicators include availability, fault tolerance, recovery capability, etc.; security indicators include confidentiality, integrity, authentication, etc.; user experience indicators include ease of use, learnability, satisfaction, etc.

[0088] Compared with the single QoS indicator of the existing method, the multi-dimensional indicator of the present invention can more comprehensively reflect the performance and quality of Web services.

[0089] 3) Evaluation model establishment: According to OWL-S semantic description and multi-dimensional evaluation indicators, an evaluation model based on OWL-S is established. The evaluation model can be constructed using AHP or fuzzy comprehensive evaluation method.

[0090] Among them, taking the AHP hierarchical analysis method to build an evaluation model as an example, it specifically includes the following steps:

[0091] 3-1) Establish a hierarchical structure: including the goal layer, criterion layer and indicator layer; the goal layer is used to calculate the comprehensive score, the criterion layer is used to evaluate functionality, performance, reliability, etc., and the indicator layer is used to set specific indicators, including functional integrity, response time, availability, etc.

[0092] 3-2) Construct a judgment matrix: compare the elements of each level in pairs to determine their relative importance; and assuming that there are n evaluation indicators, construct an n×n judgment matrix A, where element a ij Indicates the importance of indicator i relative to indicator j;

[0093] 3-3) Calculate weights: Use the eigenvector method to calculate the weights of each indicator; specifically:

[0094] 3-3-1) Calculate the maximum eigenvalue λ of the judgment matrix max and its corresponding eigenvector α;

[0095] 3-3-2) Normalize the feature vector α to get the weight vector of each indicator:

[0096] w=(w1,w2,...,w n ),

[0097] in, γ=1,…,n;w1,w2,…,w n They represent the weight of each indicator in the evaluation model; n represents the total number of evaluation indicators.

[0098] 3-4) Consistency check: ensure the consistency of judgment;

[0099] Calculate the consistency index CI and consistency rate CR:

[0100]

[0101]

[0102] Among them, RI is the random consistency index; when CR<0.1, the judgment matrix has satisfactory consistency.

[0103] 4) Evaluation result calculation: The evaluation model is used to evaluate and rank Web services, and a comprehensive score is calculated to reflect the overall performance and quality of the Web services.

[0104] 5) Machine learning optimization: Use machine learning algorithms to optimize and adjust the evaluation model, build a dynamic evaluation model, and improve the accuracy and reliability of the evaluation results. Machine learning algorithms include SVM support vector machine or random forest algorithm.

[0105] Taking the optimization of the evaluation model using SVM as an example, the specific steps include:

[0106] 5-1) Data preparation: Collect historical evaluation data, including the values ​​of each indicator and the final score;

[0107] 5-2) Feature selection: Use principal component analysis (PCA) to select the most important features;

[0108] 5-3) Model training: Use the training data set to train the SVM model; assuming that the training data set is where x i is the feature vector, yi is the score, and m is the number of training samples.

[0109] The SVM model is trained by optimizing the following objective function:

[0110]

[0111] Subject to y i (w T x i +b)≥1-ξ i ,ξ i ≥0,i=1,...,m

[0112] Among them, w is the weight vector, w T represents the transposed matrix of w, b is the bias, ξ i is the slack variable, C is the penalty parameter;

[0113] 5-4) Model validation: Use the validation dataset to evaluate model performance;

[0114] 5-5) Model application: Use the trained model to evaluate new web services and output a comprehensive score.

[0115] Compared with the existing methods, the evaluation results of the present invention are more objective and fair, and can fully reflect the performance and quality of Web services. The use of OWL-S semantic description language, multi-dimensional evaluation indicators and machine learning algorithms can provide more accurate and reliable evaluation results.

[0116] 6) Dynamic evaluation model based on machine learning: Based on user feedback and real-time data on service quality changes, the dynamic evaluation model is used to automatically adjust the evaluation index weights and dynamically evaluate and rank Web services. This model can automatically adjust the evaluation index weights based on real-time data such as user feedback and service quality changes, realize dynamic evaluation and ranking of Web services, reduce the cost and time of manual evaluation, and improve the level of automation of evaluation.

[0117] Taking the Q-learning reinforcement learning algorithm to build a dynamic evaluation model as an example, it includes:

[0118] 6-1) State definition: define the indicator values ​​of the current Web service as the state vector s;

[0119] 6-2) Action definition: Adjust the weight of each indicator as action a;

[0120] 6-3) Reward function: Define reward r based on user feedback and service quality changes;

[0121] 6-4) Q value update: Continuously update the Q value based on state-action-reward:

[0122]

[0123] Among them, α is the learning rate, which controls the influence of new information on the Q value update; γ is the discount factor, which is used to balance the immediate reward and long-term benefits; s′ is the new state after the action is performed; a′ represents the learning rate after the action is performed; Q(s,a) is the Q value of taking action a in state s, which represents the estimation of long-term benefits; Represents the maximum Q value of all possible actions in the new state s'.

[0124] 6-5) Strategy selection: Use the ε-greedy strategy to select the optimal action;

[0125] The ε-greedy strategy includes: randomly selecting an action with probability ε, that is, exploration; selecting the action with the largest current Q value with probability 1-ε, that is, utilization; the formula is expressed as:

[0126] The probability P(a) of choosing action a is:

[0127]

[0128] Among them, |A| is the size of the action space, and Q(s,a′) is the Q value of action a′ in state s.

[0129] 7) Cross-domain knowledge integration: Integrate Web service evaluation with knowledge in fields such as medicine and finance to provide professional and accurate evaluation for Web services in specific professional fields.

[0130] 8) Privacy protection mechanism: A privacy protection mechanism is adopted during the evaluation process, such as desensitizing user privacy information and using federated learning and other technologies to ensure user privacy security.

[0131] Taking the federated learning technology as an example, a privacy protection mechanism is set up, including:

[0132] 8-1) Data decentralization: Evaluation data is decentralized and stored among various participants;

[0133] 8-2) Local training: Each participant uses local data to train a local model;

[0134] 8-3) Model aggregation: The central server aggregates the local model parameters;

[0135] 8-4) Model update: Distribute the aggregated model to all participants.

[0136] The present invention aims to provide a comprehensive, accurate and reliable Web service evaluation method. By comprehensively considering multiple dimensions such as service functions, service quality, and user experience, a comprehensive evaluation index system is constructed to provide objective and fair evaluation results, thereby solving the problem that the evaluation index of existing methods is single and the results are one-sided. The efficiency and automation level of Web service evaluation are improved. By using technologies such as OWL-S semantic description and machine learning algorithms, automatic evaluation and ranking of Web services are realized, the cost and time of manual evaluation are reduced, and the evaluation efficiency is improved. The present invention enhances the adaptability and flexibility of Web service evaluation. A dynamic evaluation mechanism is introduced to automatically adjust the evaluation index weights according to real-time data such as user feedback and service quality changes to adapt to the dynamic characteristics of Web services. At the same time, personalized evaluation schemes are provided for different fields and different users.

[0137] The present invention aims to comprehensively improve and optimize Web service evaluation in terms of comprehensiveness, accuracy, efficiency, adaptability, applicability and standardization of evaluation, provide support and guarantee for the healthy development of Web services, and ultimately serve stakeholders such as Web service providers, users and platforms. The present invention enriches the application scenarios of Web service evaluation. It provides diversified evaluation services and products for different entities such as Web service providers, service users, and service platforms to meet the evaluation needs in different scenarios.

[0138] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A method for calculating the comprehensive score of Web services based on OWL-S, characterized by: The method is implemented based on the following steps, including: 1) OWL-S semantic description: Use OWL-S semantic description language to describe Web services; 2) Multi-dimensional evaluation indicators: Select multiple evaluation indicators to conduct a comprehensive evaluation of Web services; 3) Evaluation model establishment: According to OWL-S semantic description and multi-dimensional evaluation indicators, an evaluation model based on OWL-S is established; 4) Evaluation result calculation: Use the evaluation model to evaluate and rank Web services and calculate a comprehensive score to reflect the overall performance and quality of the Web services; 5) Machine learning optimization: Use machine learning algorithms to optimize and adjust the evaluation model and build a dynamic evaluation model; 6) Dynamic evaluation model based on machine learning: Based on user feedback and real-time data on service quality changes, the dynamic evaluation model is used to automatically adjust the evaluation index weights and dynamically evaluate and rank Web services; 7) Cross-domain knowledge integration: Integrate Web service evaluation with professional domain knowledge to provide professional and accurate evaluation for Web services in specific professional fields; 8) Privacy protection mechanism: A privacy protection mechanism is adopted during the evaluation process to ensure user privacy security.

2. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1 is characterized in that: The OWL-S semantic description in step 1) includes service overview, service model, service foundation, service function, service interface and service quality; The service overview is used to describe the basic information of the service, including the service name, provider, and function description; The service model is used to describe the internal workflow and composition of the service; The service foundation is used to describe how to access the service, including communication protocols and message formats.

3. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1, characterized in that: The multidimensional evaluation indicators in step 2) include functional indicators, performance indicators, reliability indicators, safety indicators, and user experience indicators; The functional indicators include functional integrity, accuracy, and applicability; The performance indicators include response time, throughput, and resource utilization; The reliability indicators include availability, fault tolerance, and recovery capability; The security indicators include confidentiality, integrity, and authentication; The user experience indicators include ease of use, learnability, and satisfaction.

4. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1, characterized in that: In step 3), the evaluation model is constructed using AHP or fuzzy comprehensive evaluation method.

5. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 4 is characterized in that: Step 3) Use AHP to build an evaluation model, which specifically includes the following steps: 3-1) Establish a hierarchical structure: including the goal layer, the criteria layer and the indicator layer; the goal layer is used to calculate the comprehensive score, the criteria layer is used to evaluate functionality, performance, and reliability, and the indicator layer is used to set specific indicators, including functional integrity, response time, and availability; 3-2) Construct a judgment matrix: Compare the elements of each level in pairs to determine their relative importance; and assuming that there are n evaluation indicators, construct an n×n judgment matrix A, where element a ij Indicates the importance of indicator i relative to indicator j; 3-3) Calculate weights: Use the eigenvector method to calculate the weights of each indicator, including: 3-3-1) Calculate the maximum eigenvalue λ of the judgment matrix max and its corresponding eigenvector α; 3-3-2) Normalize the feature vector α to get the weight vector of each indicator: in=(in1,in2,...,in n ), in, γ=1,…,n;w1,w2,…,w n Respectively represent the weight of each indicator in the evaluation model; 3-4) Consistency check: ensure the consistency of judgment; Calculate the consistency index CI and consistency rate CR: Among them, RI is the random consistency index; when CR<0.1, the judgment matrix has satisfactory consistency.

6. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1, characterized in that: In step 5), the machine learning algorithm includes SVM support vector machine or random forest algorithm.

7. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 6 is characterized in that: In step 5), the optimization of the evaluation model using SVM support vector machine specifically includes: 5-1) Data preparation: Collect historical evaluation data, including the values ​​of each indicator and the final score; 5-2) Feature selection: Use principal component analysis (PCA) to select the most important features; 5-3) Model training: Use the training data set to train the SVM model; assuming that the training data set is Among them, x i is the feature vector, y i is the score, m represents the number of training samples; The SVM model is trained by optimizing the following objective function: Subject to y i (w T x i +b)≥1-ξi ， ξ i ≥0,i=1,...,m Among them, w is the weight vector, w T represents the transposed matrix of w, b is the bias, ξ i is the slack variable, C is the penalty parameter; 5-4) Model validation: Use the validation dataset to evaluate model performance; 5-5) Model application: Use the trained model to evaluate new web services and output a comprehensive score.

8. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1, characterized in that: In step 6), the Q-learning reinforcement learning algorithm is used to build a dynamic evaluation model, including: 6-1) State definition: define the indicator values ​​of the current Web service as the state vector s; 6-2) Action definition: Adjust the weight of each indicator as action a; 6-3) Reward function: Define reward r based on user feedback and service quality changes; 6-4) Q value update: Continuously update the Q value based on state-action-reward: Among them, α is the learning rate, which controls the influence of new information on the Q value update; γ is the discount factor, which is used to balance the immediate reward and long-term benefits; s′ is the new state after the action is performed; a′ represents the learning rate after the action is performed; Q(s,a) is the Q value of taking action a in state s, which represents the estimation of long-term benefits; represents the maximum Q value of all possible actions in the new state s'; 6-5) Strategy selection: Use the ε-greedy strategy to select the optimal action; The ε-greedy strategy includes: randomly selecting an action with probability ε, that is, exploration; selecting the action with the largest current Q value with probability 1-ε, that is, utilization; the formula is expressed as: The probability P(a) of choosing action a is: Among them, |A| is the size of the action space, and Q(s,a′) is the Q value of action a′ in state s.

9. The method for calculating the comprehensive score of Web services based on OWL-S according to claim 1, characterized in that: The privacy protection mechanism in step 8) uses federated learning technology, which specifically includes: 8-1) Data decentralization: Evaluation data is decentralized and stored among various participants; 8-2) Local training: Each participant uses local data to train a local model; 8-3) Model aggregation: The central server aggregates the local model parameters; 8-4) Model update: Distribute the aggregated model to all participants.

Images