A multi-stage matching method based on preference under inter-cloud environment

Abstract

The application discloses a multi-stage matching method based on preferences under an inter-cloud environment, which comprises the following steps: screening cloud service providers satisfying the hard resource requirements of cloud service consumers by a rule matching method, and saving the cloud service providers into a candidate set S0; evaluating the performance of the cloud service providers in the candidate set S0 on the soft requirements, excluding the cloud service providers not satisfying the requirements, updating the set S0, and adopting an improved MICe method and a UMAP dimension reduction technology to process the correlation evaluation of high-dimensional data; determining a Pareto optimal solution set through a multi-objective optimization technology, further identifying a cloud service provider set S1 most conforming to the requirements in the candidate set S0 by comparing the dominance relationship of the solutions; and according to the personalized weight settings of the cloud service consumers, using a weighted summation method or a fuzzy comprehensive evaluation method to calculate the matching degrees of each cloud service provider in S1, and ranking the cloud service providers according to the matching degrees, so as to determine the final service matching result.

Description

TECHNICAL FIELD

[0001] The application relates to the cloud computing technical field, in particular to a multi-stage matching method based on preference in an inter-cloud environment. BACKGROUND

[0002] In the current digital era, cloud computing has undoubtedly become the key force to promote the rapid development of information technology, and has a series of unique advantages such as no need for high initial investment, relatively low operating cost and high scalability, thereby triggering a far-reaching revolution in the information technology field. However, the rapid evolution of cloud computing inevitably brings about a series of severe challenges while bringing many opportunities. With the continuous acceleration of its development pace, the demand for cloud services presents an explosive growth trend, which is undoubtedly a great pressure on a single cloud service provider (CSP).

[0003] During a typical special event such as the Taobao "Double 11" and the Wal-Mart "Black Friday", a large number of users simultaneously rush into the platform, and the demand for cloud service resources instantaneously reaches the peak. At this time, the resource reserve and supply capacity of a single CSP are often insufficient to meet such a huge and concentrated resource demand in a short time, thereby causing the inability to provide the required key resources for users in time, and seriously affecting the user experience and the normal operation of the business. At the same time, users are often trapped in the dilemma of "platform lock-in" during the use of cloud services. Once an enterprise chooses a specific cloud service provider, the dependence of the enterprise on the platform will continue to deepen as the business on the platform develops, and a situation of excessive dependence on a single provider will gradually form. This situation not only limits the flexibility of the enterprise in the selection of cloud services, but also may cause the enterprise to have difficulty in making effective response when facing problems such as the decline of service quality and unreasonable price rise, thereby hindering the healthy development of the cloud market.

[0004] To effectively resolve these difficult problems, academia and industry have turned their attention to the in-depth exploration of cloud 2.0 architecture. In this process, a variety of cloud cooperation modes have emerged, among which InterCloud, SuperCloud, and CloudService Broker modes have attracted much attention. The core goal of these modes is to build a new sharing mode, break through the resource limitations of individual CSPs, and promote extensive and in-depth cooperation between cloud service providers, so as to achieve mutual benefit and win-win. Among these modes, JointCloud stands out with its unique architectural design and innovative concept. It innovatively introduces two key concepts: "vertical integration" and "horizontal cooperation". "Vertical integration" focuses on breaking down the communication barriers between heterogeneous cloud stack layers, enabling seamless interaction and collaboration between layers; "horizontal cooperation" focuses on promoting close cooperation between different CSPs and integrating the resource advantages of all parties.

[0005] However, despite the many promising advantages and broad application prospects of JointCloud, it still inevitably faces a series of complex challenges in its actual operating environment. As JointCloud is widely used, it attracts more and more consumers to join in, which directly leads to an exponential increase in the variety and complexity of consumer demand. From the perspective of consumers, different consumers have significantly different preferences. Some consumers consider the reputation of the provider as the primary consideration when choosing cloud services, believing that good reputation is an important guarantee for service quality and stability; while another part of consumers are more concerned about specific attributes of cloud services, such as data processing capabilities, security levels, customization levels, etc., which are directly related to whether their business can be efficiently and smoothly carried out. From the perspective of providers, as the market continues to expand, the variety of resources available for selection is increasing, and how to accurately filter out services that perfectly match the diverse needs of consumers from this complex ocean of resources has become a highly challenging task.

[0006] In the field of service matching, although many existing methods attempt to solve the matching problem between consumers and service providers to some extent, they all have different degrees of limitations. Methods based on multi-criteria decision making (MCDM) attempt to consider multiple criteria such as cost, performance, security, etc. to rank service providers, but in practical applications, they face the difficulty of accurately determining the weights of each criterion, and they are not good at handling subjective user preferences. Matching methods based on trust and reputation mainly select cloud services by evaluating the reliability of providers, but this method often focuses too much on the reputation of providers and ignores the user's personalized preferences and specific resource requirements, which may result in the selected service having good reputation but not meeting the user's actual business requirements. Matching methods based on quality of service (QoS) focus on parameters such as availability, response time, and cost, and use optimization techniques to meet user-defined QoS constraints, but in practical operation, they are difficult to fully capture the subtle differences in user preferences for different resources and data correlations, and cannot achieve truly accurate matching. Preference-based matching mechanisms attempt to improve user satisfaction to some extent, but often ignore the user's specific resource requirements and lack effective strategies for handling whether the user provides preference information, resulting in less than ideal matching results. Utility maximization-based matching methods evaluate overall utility by assigning weights to each criterion, but also ignore the key factor of user preferences, making the evaluation results likely to deviate significantly from the user's actual expectations. Uncertainty-based matching methods use fuzzy techniques and reference point methods to deal with uncertainty in cloud service selection, but do not fully consider user preferences in the process and are prone to overfitting historical preference data, which means that when market conditions or user needs change, this method may not be able to accurately identify new services that are more beneficial to users in a timely manner.

[0007] In the JointCloud environment, there is an urgent need to design a new and more effective matching method that can accurately grasp the user's diverse preferences while fully considering resource requirements, thereby achieving more efficient and accurate matching between consumers and cloud service providers, and promoting the sustainable and healthy development of the cloud service market under the JointCloud architecture. SUMMARY

[0008] To solve the above technical problems, the purpose of the present application is to provide a multi-stage matching method based on preferences in the inter-cloud environment.

[0009] The present application provides a multi-stage matching method based on preferences in the inter-cloud environment, comprising:

[0010] Step 1: Filter out cloud service providers that meet the hard resource requirements of cloud service consumers through a rule matching method and save them to the candidate set S0;

[0011] Step 2: Evaluate the performance of cloud service providers in the candidate set S0 on soft requirements, exclude cloud service providers that do not meet the requirements, update the set S0, and the soft requirements include cost effectiveness and data resource correlation; improved MICe method and UMAP dimension reduction technology are used to process high-dimensional data correlation evaluation;

[0012] Step 3: Determine the Pareto optimal solution set through multi-objective optimization technology, and further identify the cloud service provider set S1 that best meets the requirements in the candidate set S0 by comparing the dominance relationship of solutions;

[0013] Step 4: According to the personalized weight setting of the cloud service consumer, use the weighted summation method or fuzzy comprehensive evaluation method to calculate the matching degree of each cloud service provider in S1, and rank them to determine the final service matching result.

[0014] The multi-stage matching method based on preference in the inter-cloud environment of the application mainly focuses on the inherent attributes of the service itself, and often ignores the problem of user preference. A multi-stage matching strategy is used to integrate user preference, provider reputation and data correlation to achieve more accurate matching. Rule-based matching is used to narrow the search space, reverse cost algorithm is used for scoring, and MICe index is used to evaluate data similarity. The Pareto frontier is determined by the advantage-based method, and the weighting technology customized for user preference is also used. Compared with existing technologies, the method of the application can integrate user preference, service provider reputation and data resource correlation to ensure more accurate and efficient matching in the inter-cloud environment. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 The flowchart of the multi-stage matching method based on preference in the inter-cloud environment of the application. DETAILED DESCRIPTION

[0016] This invention proposes a preference-based multi-stage matching method in a cloud environment, aiming to improve matching efficiency, quickly eliminate unqualified cloud service providers, achieve personalized matching, comprehensively consider user preferences, service provider reputation, and data resource relevance, reduce computational complexity, improve processing speed by utilizing the Maximum Information Coefficient Estimator (MICe) and dimensionality reduction techniques, enhance dynamic adaptability, flexibly respond to changes in consumer demand, strengthen privacy protection by implementing an inverse cost method to avoid user disclosure of sensitive information, and provide a comprehensive evaluation by employing a coverage decision method and entropy-weighted fuzzy comprehensive evaluation to comprehensively assess the performance of candidate cloud service providers. Through these objectives, this invention aims to provide an efficient, accurate, personalized, and user-friendly cloud service matching solution to improve the quality of cloud service selection and user satisfaction, while ensuring user privacy.

[0017] The present invention relates to a multi-stage matching method based on preferences in a cloud environment, which involves the following three main roles: cloud service provider, cloud service consumer, and cloud platform.

[0018] Cloud service providers are primarily responsible for publishing information on available resources, including computing, storage, networking, and data, as well as information on the quality of their historical services on the cloud platform.

[0019] Cloud service consumers: Select and use the most suitable cloud services on the cloud platform according to their personal needs and preferences;

[0020] Cloud Platform: Collects, processes, and stores supply and demand information published by cloud service providers and consumers. Recommends cloud service providers based on the supply and demand relationship between cloud service providers and consumers.

[0021] The interaction between the three roles is as follows: cloud service providers and cloud service consumers publish supply and demand information to the cloud platform. The cloud platform receives the information and evaluates and ranks the cloud service providers according to the supply and demand relationship. It then returns the results to the cloud service providers and cloud service consumers. Next, the cloud service providers provide the requested resources to the cloud service consumers.

[0022] like Figure 1 As shown, a preference-based multi-stage matching method in a cloud environment according to the present invention includes:

[0023] Step 1: Select cloud service providers that meet the hard resource requirements of cloud service consumers using rule matching methods and save them to the candidate set S0. Specifically:

[0024] Step 1.1: Define the hard resource requirements of cloud service consumers as:

[0025] U=(U compute U networkU storage )

[0026] where U compute , U network and U storage represent the requirements of computing, network and storage resources respectively; the resource provision vector of cloud service providers is represented as P a = (P a,compute , P a,network , P a,storage ), where P a,compute represents the amount of computing resources of the a-th cloud service provider, P a,nework represents the amount of network resources of the a-th cloud service provider, and P a,storage represents the amount of storage resources of the a-th cloud service provider.

[0027] Step 1.2: According to the requirements of cloud service consumers, define rules:

[0028] R = (R compute , R network , R storage )

[0029] where R compute , R network and R storage represent the matching rules of computing, network and storage resources respectively; the matching rules are defined as:

[0030]

[0031] where min and max represent the minimum and maximum acceptable computing resources respectively, and max represent the minimum and maximum acceptable network resources respectively, and

[0032] Step 1.3: Collect the requirements of cloud service consumers and the resource provision capabilities of cloud service providers as inputs, check whether the resources of each cloud service provider meet the rules according to the submission order, if any resource does not meet the rules, adopt the early termination strategy and immediately check the next cloud service provider; if all resources meet the rules, add the cloud service provider to the set S0.

[0033] Step 1.4: If the set is empty, output a message that no cloud service provider meets the requirements; otherwise, return the set S0, which contains cloud service providers that meet the hard resource requirements.

[0034] Step 2: Evaluate the performance of cloud service providers in the candidate set S0 on soft requirements, excluding cloud service providers that do not meet the requirements, and update the set S0. Soft requirements include cost-effectiveness and data resource relevance. The improved MICe method and UMAP dimensionality reduction technique are used to process high-dimensional data relevance evaluation to reduce computational complexity, specifically:

[0035] Step 2.1: When selecting a cloud service provider, the revenue of the cloud service consumer is a key consideration factor. The revenue formula of the cloud service consumer is defined as:

[0036] Pro = Pri - Cost

[0037] Where Pro represents revenue, Pri represents sales price, and Cost represents cost. In actual situations, CSCs usually pursue higher revenue, but not all cloud service consumers CSCs are willing to disclose the total income of their projects. Since the sales price of the cloud service consumer is fixed and only known by itself, the reverse cost scoring method is used to define the profit score. For the a-th cloud service provider, its cost is Cost a , then the corresponding reverse cost score, i.e., the profit score, is: This means that the lower the cost, the higher the reverse cost score, and the higher the potential profit score.

[0038] Step 2.2: Compare the revenue of different cloud service providers, and normalize the cost score according to the following formula to make it within the range of 0 to 1:

[0039]

[0040] Where C' a represents the normalized cost score of the a-th cloud service provider, C a represents the cost score of the a-th cloud service provider, C min and C max represent the minimum and maximum cost scores among all cloud service providers, respectively.

[0041] Through this formula, the reverse cost score of each cloud service provider CSP is mapped to a unified scale. Finally, the normalized reverse cost score C' a is used as the final revenue score Proi, so that the revenue brought by different CSPs to the CSC can be evaluated and compared under the same standard;

[0042] Step 2.3: Evaluate the relevance between the data provided by the cloud service provider and the cloud service consumer task, and record the cloud service consumer task data as Record the data provided by the cloud service provider as Each data point and is a high-dimensional vector that can comprehensively describe the data characteristics of a specific cloud service. For example, it may contain information on the scale of the data, the type distribution of the data, the access frequency of the data, and other dimensions.

[0043] Step 2.4: Select cosine similarity As a measure of the degree of similarity between data points, it is converted into cosine distance for subsequent calculation convenience, the formula is:

[0044]

[0045] Cosine distance can more intuitively reflect the difference between data points. The larger the value, the greater the difference between data points; the smaller the value, the more similar the data points.

[0046] Step 2.5: Efficiently search for the nearest neighbor of the data point in the high-dimensional space, add each data point to the Annoy index, and Annoy approximates the distribution of data points by generating multiple random projection trees. Each projection tree divides and organizes data points from different angles, enabling quick positioning of similar data points when searching for the nearest neighbor, thereby greatly improving search efficiency.

[0047] Step 2.6: For each data point find its k nearest neighbors in the Annoy index, and calculate the local scale parameter ρ i according to the following formula:

[0048]

[0049] where ρ i represents the distance between data point and its nearest neighbor; represents the k nearest neighbors of data point . ρ i can reflect the density and distribution of data points in the local range, and plays a key role in adjusting the similarity calculation in the subsequent similarity calculation, ensuring that the similarity calculation can adapt to the local characteristics of different data points.

[0050] Step 2.7: Use the calculated local scale parameter ρ i and cosine distance cosine distance (x i , x j ) to calculate the smooth distance d ij between data points and , the formula is: ​

[0051]

[0052] The calculation of the smooth distance can adjust the original cosine distance, so that the distance measurement is more smooth and continuous, and the distance mutation caused by uneven distribution of data points or noise and other factors is reduced, so as to more accurately reflect the real relationship between data points.

[0053] Step 2.8: Calculate the similarity p ij between each data point and other data points by performing exponential transformation on the smooth distance d ij , and the formula is:

[0054]

[0055] Where σ i is a smoothing parameter, which is solved by the following formula:

[0056]

[0057] The greater the value of the similarity p ij , the higher the similarity between the data points and , and in this way the local neighborhood relationship between data points in high-dimensional space can be effectively quantified.

[0058] Step 2.9: Combine the similarity between each data point and other data points to construct a high-dimensional neighborhood graph. This high-dimensional neighborhood graph can intuitively show the similarity structure between data points, providing a basis for subsequent analysis and processing.

[0059] Step 2.10: In order to reduce the computational complexity while preserving the neighborhood structure relationship between data points, dimensionality reduction processing needs to be performed on the high-dimensional data, randomly initializing the data points Y in the low-dimensional space, and then calculating the similarity between the data points and in the low-dimensional space according to the following formula:

[0060]

[0061] Use gradient descent method to minimize the cross-entropy loss function between high-dimensional and low-dimensional representations:

[0062]

[0063] Through continuous iteration optimization, the low-dimensional representation can preserve as much neighborhood structure information of the high-dimensional data as possible, thereby reducing the difficulty and computational cost of data processing without losing too much information. Finally, the data in the low-dimensional space obtained by dimensionality reduction of the cloud service consumer task data in the high-dimensional space is represented as Y A, the data representation of the low-dimensional space obtained after dimensionality reduction of the data provided by the cloud service provider is Y B .

[0064] Step 2.11: MICe is a statistical measure used to evaluate the correlation between two random variables Y A and Y B , which is obtained by calculating the sample characteristic matrix M(D) and the maximum mutual information value. For a sample set D based on the random variable pair (Y A , Y B ), the sample characteristic matrix M(D) is defined, and its elements are represented as follows:

[0065]

[0066] where I * (D, m, l) is the maximum mutual information value on the m*l grid, and m and l represent the number of rows and columns of the grid, respectively. The MICe value is obtained by summing the elements of the sample characteristic matrix, i.e.:

[0067] MICe(D) = max m,l M(D) m,l

[0068] The higher the MICe value, the stronger the average mutual information between the variables Y A and Y B , which means the stronger the correlation between them.

[0069] Step 2.12: Since MICe calculation involves calculating mutual information on a fixed grid size, the computational complexity is high. To improve computational efficiency, Bayesian optimization is used to search for the optimal grid configuration, and the objective function is defined as:

[0070] f(m, l) = -MICe(Y A ; Y B )

[0071] Because Bayesian optimization is usually a minimization problem, the reciprocal of the MICe value is taken as the objective function, and the goal is to find the grid configuration that maximizes the MICe value by minimizing this objective function.

[0072] Set the search space, choose the grid parameters m*l as the optimization variables, and determine the range and granularity of the search.

[0073] First, use Gaussian process to construct the prior distribution, calculate the initial MICe value, i.e. calculate the MICe value on the cloud service consumer and cloud service provider data set under the initial grid partition parameters.

[0074] Observe the data After that, the prior distribution is updated to the posterior distribution, considering the existence of Gaussian noise The parameters of the joint distribution and the posterior distribution are calculated.

[0075] Finally, the next sampling point is selected using the expected improvement strategy, and the expected improvement value is calculated:

[0076]

[0077] where f(m, l + ) is the current optimal solution, the expected value is calculated through the posterior distribution, and the input point that maximizes the expected improvement value is selected as the new sampling point.

[0078] Step 2.13: Repeat step 2.12 to continuously update the grid parameters and calculate the MICe value until the convergence condition is met, i.e., the change in the objective function value is less than the set threshold ∈, at which point the obtained grid configuration is the optimal configuration, thereby improving the efficiency and accuracy of MICe calculation.

[0079] Step 2.14: According to the demand of cloud service consumers for benefit score and data correlation, further filter the set S0 obtained in step 1, exclude cloud service providers that do not meet the demand, and update the set S0.

[0080] Step 3: Determine the Pareto optimal solution set, i.e., the Pareto front, through multi-objective optimization technology, and further identify the cloud service provider set S1 that best meets the requirements in the candidate set S0 by comparing the dominance relationship of solutions, specifically:

[0081] Step 3.1: In the cloud environment, in order to comprehensively evaluate cloud service providers, a multi-objective function is defined:

[0082] min f(X) = (-f1(X), -f2(X), …, -f6(X))

[0083] Here is a decision vector variable, where X af1(X) represents the amount of computing resources provided by the cloud service provider, which is a key indicator of whether the CSP can meet the CSC's computing task requirements; f2(X) represents the amount of network resources, which relates to the speed and stability of data transmission; f3(X) is the amount of storage resources, which determines the size of the data that the cloud service provider can store; f4(X) is the profit score, reflecting the economic value of the cloud service provider to the cloud service consumer; f5(X) embodies data relevance, meaning the degree of association between the data provided by the cloud service provider and the task of the cloud service consumer; f6(X) is the reputation score, representing the reputation and reputation of the cloud service provider in the market. Through such a comprehensive objective function, the performance of the CSP in multiple dimensions can be comprehensively evaluated to screen out the most suitable CSP for the CSC.

[0084] The candidate set S0 is composed of the remaining CSPs after the first two stages (the first stage of rule-based matching screening and the second stage of soft demand calculation). The first stage has ruled out those CSPs that do not meet the hard resource requirements through rule-based matching, and the second stage has calculated the soft demand information among the remaining CSPs. After the screening of these two stages, the remaining CSPs meet some of the requirements of the CSC to some extent, but there may still be some redundant or with room for improvement in multi-objective optimization. These remaining CSPs together constitute the candidate set of this stage, waiting for further screening and optimization.

[0085] Step 3.2: Each solution X of the candidate set S0 a needs to be compared with each solution X p of the current Pareto front P to determine their dominance relationship; if X a is not inferior to X p in some aspects and is superior to X p in at least one objective function, then X a dominates X p , denoted as X a < X p ; assume X a is not dominated, marked as dominated = false; traverse each X p in the Pareto front P, if X p dominates X a , then immediately mark dominated as true and stop comparing X p , because it has been determined that X a is dominated; conversely, if X a dominates X p , then X pAdd it to the to-remove list because the dominated solution will be removed when the Pareto front is updated later.

[0086] Step 3.3: After completing a solution X in the candidate set S0 a After determining the dominance relationship with all solutions in the Pareto front P, if X a If the solution is not dominated, then the Pareto front P needs to be updated by iterating through each solution X in the column to be removed, to_remove. r Remove them from P because these solutions have been replaced by the new solution X. a Controlled. X a Adding these solutions to P, which now contains all the undominated solutions so far, forms a new Pareto front. By continuously updating the Pareto front, the final matching set can include the solution of the best cloud service provider in the sense of multi-objective optimization. That is, these solutions achieve a relatively optimal balance among the various objective functions, and no other solution can improve the value of a certain objective function without decreasing the values ​​of other objective functions.

[0087] Step 3.4: After traversing all solutions in the candidate set S0, the Pareto front P obtained at this time is the final matching set. This matching set contains the set S1 of cloud service providers that are most suitable for cloud service consumers, selected according to the defined objective function and the judgment dominance relationship under the multi-objective optimization framework. Output this Pareto front P to provide a basis for subsequent decision-making. Cloud service consumers can select the cloud service provider that best meets their needs and preferences to cooperate with based on this matching set, thereby obtaining a better experience and benefits in cloud computing services.

[0088] Step 4: Based on the personalized weight settings of cloud service consumers, calculate the matching degree of each cloud service provider in S1 using a weighted summation method or a fuzzy comprehensive evaluation method, and rank them accordingly to determine the final service matching result. Specifically:

[0089] Step 4.1: In the cloud computing environment, each evaluation index dimension such as computing, network and storage resources is diverse and the data is unevenly distributed, but the interval is relatively fixed. In order to make these indicators comparable, according to the specific needs of cloud service consumers, the upper and lower limits of each cloud service provider's resource indicators, including computing resource amount, network resource amount and storage resource amount, are explicitly set. For example, for computing resources, if the CSC requires processing capacity within a certain range, the upper and lower limits of the computing resource indicator can be determined accordingly. The Min-Max normalization method is used to process the resource indicators. When the maximum and minimum values of the indicators are different, the resource indicator values are mapped to the interval of 0 to 1. If the maximum and minimum values are equal, the indicator value is set to 1. In this way, indicators with different ranges and distributions can be converted into a unified standard for subsequent calculation and comparison. The data is already standardized, so only the reputation score of the cloud service provider needs to be standardized to ensure that all indicators used to evaluate the cloud service provider are under the same measurement standard.

[0090] Step 4.2: When the cloud service consumer clearly expresses the importance of each evaluation indicator, the optimization expert scoring method is used to invite experts in the relevant field to score each indicator. These experts, based on their professional knowledge and experience, evaluate the performance of cloud service providers in each indicator. Experts may score cloud service providers based on computing resource stability, network resource transmission speed, and storage resource security. Then, through normalization processing, the weight of each evaluation indicator is determined, and the formula is:

[0091]

[0092] where x cd is the standardized value of the dth evaluation indicator of the cth cloud service provider, ω cd represents the weight of the dth evaluation indicator of the cth cloud service provider, and N is the total number of evaluation indicators. The sum of the weights of each evaluation indicator for each service provider is 1, and the weight allocation accurately reflects the preferences of the cloud service consumer. Evaluation indicators include computing resource amount, network resource amount, storage resource amount, profit score, data relevance and reputation score.

[0093] The final score is calculated according to the weight:

[0094]

[0095] H d represents the final score of the dth cloud service provider; X cd refers to the normalized cth evaluation indicator of the dth cloud service provider. According to the final score, all cloud service providers are ranked.

[0096] Step 4.3: If the cloud service consumer does not disclose the importance of each indicator, in order to objectively determine the weight, the entropy weight method is used to weight each evaluation indicator, and a normalized matrix Z is constructed based on the previously standardized data, and the Z cd Element represents the standardized data of the cth cloud service provider on the dth indicator, and the information entropy e of each indicator is calculated according to the definition of entropy d :

[0097]

[0098] Where M is the number of cloud service providers, and the information entropy reflects the dispersion degree of the indicator data. The greater the information entropy, the smaller the amount of information provided by the indicator, and the smaller the weight. Conversely, the smaller the information entropy, the greater the weight.

[0099] The information entropy is calculated according to the following formula to calculate the weight of each evaluation indicator:

[0100]

[0101] Step 4.4: Use fuzzy comprehensive evaluation method to define membership function. Since the data distribution of each indicator has uncertainty and fuzziness, it is necessary to calculate relevant statistics such as mean, variance, etc. to understand the concentration trend and dispersion degree of the data. According to these statistical characteristics, different types of membership functions are fitted, such as triangular membership function, trapezoidal membership function, etc. By comparing the fitting degree of each membership function and the data distribution, the function with the highest fitting degree is selected as the final membership function; the indicator value of the cloud service provider can be converted into fuzzy membership value, and a comprehensive evaluation matrix R is constructed, where element r cd represents the fuzzy membership value of the cth cloud service provider on the dth indicator.

[0102] Step 4.5: Calculate the fuzzy comprehensive evaluation value of each cloud service provider by fuzzy weighted sum, and the formula is:

[0103]

[0104] S c represents the fuzzy comprehensive evaluation value of the cth cloud service provider, and w d is the weight of the dth average indicator. The fuzzy membership value of each indicator of each cloud service provider is weighted and summed with the corresponding weight to obtain an evaluation value that comprehensively reflects the overall performance of the cloud service provider.

[0105] Step 4.6: According to the calculated fuzzy comprehensive evaluation value, all cloud service providers in S1 are ranked from high to low, and the ranked cloud service provider list can clearly show the order of the advantages and disadvantages of each cloud service provider in meeting the needs of cloud service consumers, thereby facilitating the cloud service consumers to make the optimal choice according to their own needs and select the cloud service provider that best meets their expectations for cooperation.

[0106] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. The present application can be variously changed and modified by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A preference-based multi-stage matching method in a cloud environment, characterized in that, include: Step 1: Select cloud service providers that meet the hard resource requirements of cloud service consumers using rule matching methods and save them to the candidate set S0; Step 2: Evaluate the performance of cloud service providers in candidate set S0 on soft requirements, eliminate cloud service providers that do not meet the requirements, and update set S0. Soft requirements include cost-effectiveness and relevance of data resources. An improved MICe method and UMAP dimensionality reduction technique are used to process the correlation assessment of high-dimensional data; Step 3: Determine the Pareto optimal solution set through multi-objective optimization techniques, and further identify the set of cloud service providers S1 that best meet the requirements from the candidate set S0 by comparing the dominance relationships of the solutions; Step 4: Based on the personalized weight settings of cloud service consumers, use the weighted summation method or fuzzy comprehensive evaluation method to calculate the matching degree of each cloud service provider in S1, and rank them accordingly to determine the final service matching result; Step 2 specifically involves: Step 2.1: When choosing a cloud service provider, revenue is a key consideration for cloud service consumers. The revenue formula for cloud service consumers is defined as follows: Pro = Pri-Cost Where Pro represents revenue, Pri represents selling price, and Cost represents cost; Since the selling price of cloud services to consumers is fixed and known only to themselves, a reverse cost scoring method is used to define the profit score. For the a-th cloud service provider, its cost is Cost. a The corresponding reverse cost score, i.e., the profit score, is: This means that the lower the cost, the higher the reverse cost score, and the higher the potential profit score; Step 2.2: Compare the revenue of different cloud service providers, and normalize the cost scores according to the following formula to make them fall within the range of 0 to 1: Among them, C' a C represents the normalized cost score of the a-th cloud service provider. a C represents the cost score of the a-th cloud service provider. min and C max These represent the minimum and maximum cost scores among all cloud service providers, respectively. Step 2.3: Evaluate the correlation between the data provided by the cloud service provider and the cloud service consumer task, and denote the cloud service consumer task data as... Data provided by cloud service providers is denoted as Each data point and These are all high-dimensional vectors, which can comprehensively describe the data characteristics of a specific cloud service; Step 2.4: Select cosine similarity As a metric for measuring the similarity between data points, it is converted to cosine distance for ease of subsequent calculations, with the formula: cosine distance It can more intuitively reflect the degree of difference between data points. The larger the value, the greater the difference between data points; the smaller the value, the more similar the data points are. Step 2.5: Efficiently search for the nearest neighbors of data points in high-dimensional space and add each data point to the Annoy index. Annoy approximates the distribution of data points by generating multiple random projection trees. Step 2.6: For each data point Find its k nearest neighbors in the Annoy index, and calculate the local scale parameter ρ according to the following formula. i : Where, ρ i Representing data points The distance to its nearest neighbor; Representing data points k nearest neighbors; Step 2.7: Utilize the calculated local scale parameter ρ i and cosine distance cosine distance (x i ,x j To calculate data points and Smooth distance d ij The formula is: Step 2.8: By adjusting the smoothing distance d ij Perform an exponential transformation to calculate the similarity p ij The formula is: Where, σ i It is a smoothing parameter, obtained by solving the following formula: Step 2.9: Combine the similarity between each data point and other data points to construct a high-dimensional neighborhood graph; Step 2.10: To reduce computational complexity while preserving the neighborhood structure relationships between data points, it is necessary to perform dimensionality reduction on the high-dimensional data. Data points Y are randomly initialized in the low-dimensional space, and then the data points in the low-dimensional space are calculated according to the following formula. and Similarity between them: Minimize the cross-entropy loss function between the high-dimensional and low-dimensional representations using gradient descent: Through continuous iterative optimization, the low-dimensional representation can retain as much neighborhood structure information as possible from the high-dimensional data, thereby reducing the difficulty and computational cost of data processing without losing too much information. Ultimately, the low-dimensional representation of the cloud service consumer task data from the high-dimensional space is represented as Y. A The data provided by the cloud service provider, after dimensionality reduction, is represented in a low-dimensional space as Y. B ; Step 2.11: MICe is a method for evaluating two random variables Y. A and Y B The correlation statistic between them is obtained by calculating the MICe value through the sample feature matrix M(D) and the maximum mutual information value; for random variable pairs (Y... A ,Y B Let D be a sample set, and define the sample feature matrix M(D), whose elements are represented as follows: Among them, I * (D,m,l) is the maximum mutual information value on an m*l grid, where m and l represent the number of rows and columns of the grid, respectively; the MICe value is obtained by summing the elements of the sample feature matrix, i.e.: MICe(D)=max m,l M(D) m,l The higher the MICe value, the better the variable Y. A and Y B The stronger the average mutual information between them, the stronger the correlation between them. Step 2.12: Use Bayesian optimization to search for the optimal grid configuration, defining the objective function as: f(m,l)=-MICe(Y A ;Y B ) Set the search space, select the grid parameter m*l as the optimization variable, and determine the search range and granularity; First, a prior distribution is constructed using a Gaussian process, and the initial MICe value is calculated. That is, under the initial grid partitioning parameters, the MICe value on the cloud service consumer and cloud service provider datasets is calculated. In observing data Then, the prior distribution is updated to the posterior distribution, taking into account the presence of Gaussian noise. Calculate the parameters of the joint distribution and the posterior distribution; Finally, the expected improvement strategy is used to select the next sampling point, and the expected improvement value is calculated: Where, f(m,l) + ) is the current optimal solution. The expected value is calculated through the posterior distribution, and the input point that maximizes the expected improvement value is selected as the new sampling point. Step 2.13: Repeat step 2.12 to continuously update the grid parameters and calculate the MICe value until the convergence condition is met, that is, the change in the objective function value is less than the set threshold ∈. The grid configuration obtained at this time is the optimal configuration, thereby improving the efficiency and accuracy of MICe calculation. Step 2.14: Based on the cloud service consumers' needs for benefit scores and data relevance, further filter the set S obtained in Step 1, exclude cloud service providers that do not meet the needs, and update the set S.

2. The preference-based multi-stage matching method in a cloud environment according to claim 1, characterized in that, Step 1 specifically involves: Step 1.1: Define the hard resource requirements of cloud service consumers as: U=(U compute ,IN network ,IN storage ) Among them, U compute U network and U storage These represent the demands for computing, networking, and storage resources, respectively; the resource provision vector of the cloud service provider is represented by P. a =(P a,compute ,P a,network ,P a,storage ), where P a,compute P represents the amount of computing resources provided by the a-th cloud service provider. a,network P represents the network resource quantity of the a-th cloud service provider. a,storage This represents the amount of storage resources provided by the a-th cloud service provider; Step 1.2: Define rules based on the needs of cloud service consumers: R=(R compute ,R network ,R storage ) Among them, R compute R network and R storage These represent the matching rules for computing, network, and storage resources, respectively; the matching rules are defined as follows: in, These represent the minimum and maximum acceptable computing resources, respectively. These represent the minimum and maximum acceptable network resources, respectively. and These represent the minimum and maximum acceptable storage resources, respectively. Step 1.3: Collect the demand of cloud service consumers and the resource provision capabilities of cloud service providers as input. Check whether the resources of each cloud service provider meet the rules in the order of submission. If any resource does not meet the rules, adopt an early termination strategy and immediately check the next cloud service provider. If all resources meet the rules, add the cloud service provider to the set S. Step 1.4: If the set is empty, output the message that no cloud service provider meets the requirements; otherwise, return set S, which contains cloud service providers that meet the hard resource requirements.

3. The preference-based multi-stage matching method in a cloud environment according to claim 1, characterized in that, Step 3 specifically involves: Step 3.1: In a cloud environment, a multi-objective function is defined to comprehensively evaluate cloud service providers: minf(X)=(-f1(X),-f2(X),…,-f6(X)) here It is a decision vector variable, where X a Let f1(X) represent the decision vector of the a-th cloud service provider; f2(X) represent the amount of computing resources provided by the cloud service provider; f3(X) represent the amount of network resources, which are related to the speed and stability of data transmission; f4(X) represent the amount of storage resources, which determines the amount of data that the cloud service provider can store; f5(X) represent the profit score, which reflects the economic value of the cloud service provider to cloud service consumers; f6(X) represent the data relevance, which means the degree of correlation between the data provided by the cloud service provider and the tasks of cloud service consumers; and f6(X) represent the reputation score, which represents the reputation and word-of-mouth of the cloud service provider in the market. Step 3.2: For each solution X in the candidate set S0 a It is necessary to compare each solution X with the current Pareto front P. p Compare them to determine their dominance relationship; if for all objective functions, X a In some respects, it is not inferior to X p And it outperforms X on at least one objective function. p Then we call X a DominateX p , denoted as X a <X p Assume X a Not dominated, marked as dominated=false; traverse each X in the Pareto front P p If X is found p DominateX a If so, immediately mark dominated as true and stop dominating X. p The comparison; conversely, if X a DominateX p Then X p Add it to the to-remove list because the dominated solution will be removed when the Pareto front is updated later. Step 3.3: After completing a solution X in the candidate set S0 a After determining the dominance relationship with all solutions in the Pareto front P, if X a If the solution is not dominated, then the Pareto front P needs to be updated by iterating through each solution X in the column to be removed, to_remove. r Remove it from P; remove X a Adding to P, P contains all the undominated solutions so far, forming a new Pareto front; by continuously updating the Pareto front, the final matching set can contain the solution of the best cloud service provider in the sense of multi-objective optimization. Step 3.4: After traversing all solutions in the candidate set S0, the Pareto front P obtained at this time is the final matching set. This matching set contains the set S1 of cloud service providers that are most suitable for cloud service consumers, selected according to the defined objective function and the judgment dominance relationship under the multi-objective optimization framework. Output this Pareto front P to provide a basis for subsequent decision-making. Cloud service consumers can select the cloud service provider that best meets their needs and preferences to cooperate with based on this matching set, thereby obtaining a better experience and benefits in cloud computing services.

4. The preference-based multi-stage matching method in a cloud environment according to claim 3, characterized in that, Step 4 specifically involves: Step 4.1: Based on the specific needs of cloud service consumers, clearly define the upper and lower limits of resource indicators for each cloud service provider. Resource indicators include computing resources, network resources, and storage resources. Use the Min-Max standardization method to process the resource indicators. When the maximum and minimum values ​​of the indicators are different, map the resource indicator values ​​to the range of 0 to 1. If the maximum and minimum values ​​are equal, set the indicator value to 1. The data is already standardized; we only need to standardize the reputation scores of cloud service providers to ensure that all metrics used to evaluate cloud service providers are under the same measurement standard. Step 4.2: When cloud service consumers explicitly express their emphasis on each evaluation indicator, an optimized expert scoring method is adopted. Experts in relevant fields are invited to score each indicator. These experts, based on their professional knowledge and experience, evaluate the cloud service provider's performance on each indicator. Experts may score based on the stability of the cloud service provider's computing resources, the transmission speed of network resources, and the security of storage resources. Then, the weight of each evaluation indicator is determined through normalization, using the following formula: Where, x cd ω is the standardized value of the d-th evaluation metric for the c-th cloud service provider. cd This represents the weight of the d-th evaluation indicator for the c-th cloud service provider, where N is the total number of evaluation indicators. The sum of the weights of each evaluation indicator for each service provider is 1, and the weight allocation accurately reflects the preferences of cloud service consumers. The evaluation indicators include computing resources, network resources, storage resources, profit score, data relevance, and reputation score. The final score is calculated based on the weights: H d X represents the final score of the d-th cloud service provider; cd This refers to the d-th evaluation metric after normalization of the c-th cloud service provider, which ranks all cloud service providers based on the final score. Step 4.3: If cloud service consumers have not disclosed the importance of each indicator, in order to objectively determine the weights, the entropy weight method is used to weight each evaluation indicator. A normalized matrix Z is constructed based on the previously standardized data. cd The element represents the standardized data of the c-th cloud service provider on the d-th metric, and the information entropy e of each metric is calculated according to the definition of entropy. d : Where M represents the number of cloud service providers, and information entropy reflects the degree of dispersion of indicator data. The larger the information entropy, the smaller the amount of information provided by the indicator, and the smaller the weight should be; conversely, the smaller the information entropy, the larger the weight. The weights of each evaluation index are calculated using the following formula: Step 4.4: Define membership functions using the fuzzy comprehensive evaluation method. Due to the uncertainty and fuzziness of the data distribution for each indicator, it is necessary to first calculate relevant statistics to understand the central tendency and dispersion of the data. Based on these statistical characteristics, fit different types of membership functions. By comparing the fit between each membership function and the data distribution, select the function with the highest fit as the final membership function. This transforms the indicator values ​​of the cloud service provider into fuzzy membership values, constructing a comprehensive evaluation matrix R, where the element r... cd This represents the fuzzy membership value of the c-th cloud service provider on the d-th metric; Step 4.5: Calculate the fuzzy comprehensive evaluation value for each cloud service provider using fuzzy weighted summation. The formula is: S c w represents the fuzzy comprehensive evaluation value of the c-th cloud service provider. d It is the weight of the d-th average indicator. The fuzzy membership values ​​of each indicator of each cloud service provider and their corresponding weights are weighted and summed to obtain an evaluation value that comprehensively reflects the overall performance of the cloud service provider. Step 4.6: Based on the calculated fuzzy comprehensive evaluation value, sort all cloud service providers in S1 from high to low. The sorted list of cloud service providers can clearly show the order of merits and demerits of each cloud service provider in meeting the needs of cloud service consumers, so as to facilitate cloud service consumers to make the best choice according to their own needs and select the cloud service provider that best meets their expectations for cooperation.

Images