A combined service reconfiguration method, apparatus, device and computer readable medium

Abstract

The application discloses a kind of combination service reconfiguration method, device, equipment and computer readable medium, belong to computer technical field.The method comprises: predicting the general service quality and domain service quality of original component service in combination service;General service quality and domain service quality of original component service are compared with the general service quality constraint and domain service quality constraint corresponding to original component service respectively, and the original component service failure in combination service is determined according to the comparison result;For the failure of original component service, the corresponding replacement component service is obtained;Replacement component service is used to replace the failure of original component service, and reconfiguration combination service is generated.The technical scheme disclosed in the application comprehensively predicts the failure of general service quality (GQoS) and domain service quality (DQoS) when judging the service QoS failure, realizes dynamic reconfiguration of combination service in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to a combination service reconstruction method, device, equipment and computer readable medium. BACKGROUND

[0002] In recent years, with the emergence of large-scale business services from multiple fields on the network, business services from different fields form a service network through mutual cooperation to provide users with rich service resources. Service optimization combination technology aggregates simple services from different fields to build powerful combination services, thereby meeting the complex daily life and work business needs of users. At the same time, with the popularity of mobile networks and smart terminals, users can access services anytime and anywhere.

[0003] At present, more and more users tend to use combination services to handle various businesses. However, due to the dynamic nature of the service running environment, the quality of the combination service is affected by multiple factors, showing unstable characteristics. During the running of the combination service, the combination service often fails due to poor quality, thereby causing the combination service to fail, seriously affecting the user experience and the utilization rate of service resources, which requires the quality of the combination service to be reconstructed. In the prior art, although some technical achievements have been made in the reconstruction of combination services, there are still problems such as incomplete combination service failure judgment index and inaccurate judgment basis. SUMMARY

[0004] In order to solve at least one technical problem in the prior art described above, the embodiments of the present application provide a combination service reconstruction method, device, equipment and computer readable medium. The technical solution is as follows:

[0005] In a first aspect, a combination service reconstruction method is provided, comprising:

[0006] predicting the general service quality and the domain service quality of the original component service in the combination service;

[0007] comparing the general service quality and the domain service quality of the original component service with the general service quality constraint and the domain service quality constraint corresponding to the original component service, respectively, and determining the failed original component service in the combination service according to the comparison result;

[0008] obtaining the corresponding replacement component service for the failed original component service;

[0009] using the replacement component service to replace the failed original component service to generate a reconstructed combination service.

[0010] Further, obtaining the general service quality constraint corresponding to the original component service comprises:

[0011] obtaining a general service quality constraint corresponding to the combined service;

[0012] decomposing the general service quality constraint corresponding to the combined service into general service quality constraints corresponding to each of the original component services in the combined service by using a social learning optimization algorithm.

[0013] Further, the decomposing the general service quality constraint corresponding to the combined service into general service quality constraints corresponding to each of the original component services in the combined service by using a social learning optimization algorithm comprises:

[0014] mapping a decomposition task of decomposing the general service quality constraint corresponding to the combined service into general service quality constraints corresponding to each of the original component services in the combined service into an original individual;

[0015] respectively performing evolutionary operation and learning operation on the original individual in a micro-space and a learning space to obtain an evolved individual and a learned individual; the evolutionary operation comprises: crossover operation and / or mutation operation, and the learning operation comprises: observation learning and / or imitation learning;

[0016] calculating fitness values of the original individual, the evolved individual and the learned individual, and performing greedy operation on the original individual respectively with the evolved individual and the learned individual according to the fitness values to retain individuals with higher fitness values;

[0017] uploading the individuals with higher fitness values in the population to the belief space according to an update frequency;

[0018] replacing the individuals with lower fitness values in the micro-space by the individuals in the belief space.

[0019] Further, a case-based reasoning method is adopted to jointly predict the general service quality and the domain service quality of the original component services in the combined service, comprising:

[0020] representing the original component services as target cases, wherein condition features of the target cases comprise context information of the original component services at a time of imminent running, and decision features of the target cases comprise general service quality and domain service quality corresponding to future running of the original component services;

[0021] determining similar cases of the target cases in a historical case library according to similarity of the target cases and the historical cases;

[0022] predicting the general service quality and the domain service quality of the target cases according to the general service quality and the domain service quality in the similar cases.

[0023] Further, the determining the similar case of the target case in the historical case library according to the similarity between the target case and the historical case comprises:

[0024] An optimal weight combination corresponding to different decision features is determined by using a culture-genetic algorithm.

[0025] The similarity between the target case and the historical case is calculated according to the optimal weight combination.

[0026] The similar case is determined in the historical case library according to the similarity condition and the similarity between the target case and the historical case.

[0027] Further, the method further comprises:

[0028] At least one candidate component service is determined for the failed original component service in a component service library based on a minimum spanning tree heuristic segmentation clustering method, the component service library comprising a plurality of resource component services; wherein the method comprises:

[0029] The average similarity of the resource component services is taken as a cut-off distance to calculate the local density of each resource component service.

[0030] The resource component services are sorted according to the size of the service local density, the points corresponding to the resource component services satisfying the sorting condition are taken as initial class cluster centers, and the remaining resource component services are distributed according to the similarity.

[0031] Each class cluster is mapped as a node, the similarity between class clusters is mapped as the weight of the edge between nodes, the intra-cluster similarity and the inter-cluster similarity of each class cluster are calculated, and the kruskal algorithm is used to select the edge with the maximum weight to generate a minimum spanning tree.

[0032] The minimum weight edge in the minimum spanning tree is segmented to obtain new class clusters, and the result is stored as a new clustering result.

[0033] The segmentation operation is stopped when all edges are cut, the ratio of the intra-cluster similarity to the inter-cluster similarity of each clustering result is calculated, the ratio with the minimum value is selected as the final clustering result, and the candidate component service is obtained.

[0034] Further, the method further comprises:

[0035] When the candidate component services are multiple, the replacement component service is determined in the set of candidate component services according to the general service quality and the field service quality of the candidate component services.

[0036] In a second aspect, a combined service reconstruction device is provided, and the device comprises:

[0037] a service quality prediction module configured to predict general service quality and domain service quality of original component services in a combined service;

[0038] a failed service determination module configured to compare the general service quality and the domain service quality of the original component services with general service quality constraints and domain service quality constraints corresponding to the original component services respectively, and determine failed original component services in the combined service according to comparison results;

[0039] a replacement service determination module configured to obtain corresponding replacement component services for the failed original component services;

[0040] a combined service reconstruction module configured to replace the failed original component services with the replacement component services, and generate a reconstructed combined service.

[0041] In a third aspect, an electronic device is provided, and the device comprises:

[0042] one or more processors; and

[0043] a memory associated with the one or more processors, the memory being configured to store program instructions, which, when executed by the one or more processors, perform the method according to any one of the first aspect.

[0044] In a fourth aspect, a computer readable medium is provided, and the medium stores a computer program, wherein the program, when executed by a processor, implements the method according to any one of the first aspect.

[0045] The technical scheme provided by the embodiments of the present application has at least the following beneficial effects:

[0046] 1. The technical scheme disclosed in the present application can comprehensively predict the failure of general service quality (GQoS) and domain service quality (DQoS) of a service, and dynamically reconstruct a combined service in a timely manner when judging the service QoS failure;

[0047] 2. When obtaining the GQoS constraints corresponding to the component services, the technical scheme disclosed in the present application adopts a global GQoS constraint optimization decomposition method based on user preference perception, which decomposes the global GQoS constraints that the combined service needs to satisfy into local GQoS constraints that each component service needs to satisfy, thereby obtaining a more accurate basis for judging the GQoS failure of the component services;

[0048] 3. The technical scheme disclosed by the application adopts a minimum spanning tree heuristic segmented service clustering algorithm when searching for a replacement service for an invalid service, and performs similar service clustering on large-scale services according to the multi-dimensional attributes of the services, thereby laying an important foundation for realizing fast similar service discovery. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical schemes in the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative effort on the basis of these drawings.

[0050] Figure 1 is a combined service reconstruction method flowchart provided by the embodiment of the application;

[0051] Figure 2 is a preference-aware GQoS value interval determination process provided by the embodiment of the application;

[0052] Figure 3 is a GQoS quality scale decomposition process schematic diagram provided by the embodiment of the application;

[0053] Figure 4 is a cross operation process schematic diagram of individual X i An encoding example diagram;

[0054] Figure 5 is a cross operation process schematic diagram of individual X i and X j ;

[0055] Figure 6 is a mutation operation process schematic diagram of individual X i ;

[0056] Figure 7 is an observation learning process schematic diagram of individual X i ;

[0057] Figure 8 is an imitation learning process schematic diagram of individual X i ;

[0058] Figure 9 is a combined service reconstruction device module schematic diagram provided by the embodiment of the application;

[0059] Figure 10 is a computer system structure schematic diagram provided by the embodiment of the application. DETAILED DESCRIPTION

[0060] In order to make the objects, technical solutions and advantages of the present application clearer, 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 a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0061] In recent years, researchers in the field have carried out preliminary research on the dynamic reconstruction of composite services in the case of service failure or quality of service failure, and have achieved certain research results. For the problem of QoS failure of a single component service in a composite service, various QoS failure identification methods are proposed, and the QoS failure service is replaced. For example: a reconstruction strategy combining single service replacement and regional reconstruction is adopted, and the research focus is placed on how to quickly realize the reconstruction of the failure region of the composite service; by determining the QoS failure region in the composite service, the composite service regional reconstruction problem is converted into a service composition problem, and optimization algorithms, graph theory and integer programming methods are used to solve it. Although the existing research on dynamic reconstruction of composite services for service quality failure has achieved certain research results, there are still the following deficiencies:

[0062] (1) The existing research work does not give how to obtain accurate judgment basis for component service QoS failure when judging service QoS failure, resulting in low accuracy of service QoS failure judgment;

[0063] (2) The existing research work does not consider the case of domain quality of service (DQoS) failure when judging service QoS failure, missing some opportunities for dynamic reconstruction of composite services;

[0064] (3) When looking for replacement services for failed services, the existing service discovery method usually uses the traversal method to select suitable services from a large number of candidate services. However, due to the existence of a large number of available services on the network, the existing service discovery method is low in efficiency and long in running time, which seriously affects the timeliness of dynamic reconstruction of composite services.

[0065] In view of the above problems, as shown in the present application, a composite service reconstruction method is provided, and the specific technical solutions are as follows: Figure 1 S1, predicting the general service quality and domain service quality of the original component service in the composite service.

[0066]

[0067] ​The above-mentioned one or more original component services are contained in the combined service. The general service quality (GQoS) and the domain service quality (DQoS) of the original component service are predicted, mainly referring to the GQoS and the DQoS of the original component service in the process of executing the next running task of the combined service. The specific prediction method can be artificial experience estimation, estimation according to the GQoS and the DQoS of the combined service in the process of executing a similar running task, or estimation according to the GQoS and the DQoS of other combined services similar to the combined service in the process of executing the same or similar running task.

[0068] In one embodiment, step S1 is specifically using a case-based reasoning method to jointly predict the general service quality and the domain service quality of the original component service in the combined service.

[0069] The above-mentioned case-based reasoning (CBR) is an efficient problem solving method. A case generally consists of condition characteristics and decision characteristics. It solves new problems by storing similar problems and reusing the solutions of these problems. The CBR solving process model includes four links, namely case retrieval, case reuse, case revision and case storage. The embodiment of the present application proposes an improved case-based reasoning method, which can jointly predict the GQoS and the DQoS of the original component service by retrieving historical cases similar to the target case and according to these similar cases.

[0070] The case-based reasoning method disclosed by the embodiment of the present application specifically includes:

[0071] The original component service is represented as a target case. The condition characteristics of the target case include the context information of the original component service when it is about to run. The decision characteristics of the target case include the general service quality and the domain service quality corresponding to the original component service in the future running;

[0072] According to the similarity between the target case and the historical cases, the similar cases of the target case are determined in the historical cases;

[0073] The general service quality and the domain service quality of the target case are predicted according to the general service quality and the domain service quality in the similar cases.

[0074] The above-mentioned original component service is first represented as a case. The condition characteristics of the case are the context information of the original component service when it runs, including: the load of the service, the task type (TT), the total amount of the task (TQ), etc. The decision characteristics include the GQoS value and the DQoS value exhibited by the service when processing the task. The case model definition for GDQoS prediction is shown in formula (1):

[0075]

[0076] where Load i , TT i , TQ i denote the load, the task type and the total amount of tasks of the i-th instance, respectively. and denote the GQoS value and the DQoS value of the service when processing the task with the task type of TT i and the task amount of TQ i under the load of Load i .

[0077] The instance retrieval is the first step of the four steps of the CBR, and the retrieval quality directly affects the performance of the CBR model, and the unreasonable weight distribution will affect the retrieval quality. Therefore, it is necessary to reasonably distribute the features. The instance retrieval refers to retrieving K similar instances to the target instance from the instance library within a specified time. The embodiment of the application uses similarity to measure the closeness between the target instance and the historical instance, and then realizes the retrieval of similar cases. The target GDQoS instance is defined as:

[0078]

[0079] where Load0, TT0, TQ0 are extracted from the user's request information and the service running data; and denote the GQoS value and the DQoS value to be predicted.

[0080] The similarity between the target instance and the historical instance is divided into local similarity and global similarity. The local similarity is the similarity between different condition features in the target instance and the historical instance, and the global similarity is the weighted sum of the local similarity. In the embodiment of the application, all types of condition attributes are numerical attributes, and therefore the Hamming distance formula (3) is used to calculate the similarity of numerical attributes:

[0081]

[0082] where, denotes the p-th condition feature value of the historical instance, denotes the p-th condition feature value of the target instance, and denote the minimum value and the maximum value of the p-th condition feature of all instances. The global similarity calculation formula based on the Hamming distance is shown in formula (4):

[0083]

[0084] where, TC0 denotes the target instance, SCi sim n sim i sim n sim

[0085] Since the importance of each condition feature to different decision features is different, in an embodiment, a cultural-genetic algorithm is used to optimize the combination of the weight values corresponding to different decision features, so as to obtain more similar cases in the case retrieval link and improve the accuracy of GDQoS prediction.

[0086] Based on the above, in the embodiment of the present application, the similar cases of the target case are determined in the historical case library according to the similarity between the target case and the historical cases, comprising:

[0087] The cultural-genetic algorithm is used to determine the optimal combination of the weight values corresponding to different decision features;

[0088] The similarity between the target case and the historical cases is calculated according to the optimal combination of the weight values;

[0089] The similar cases are determined in the historical case library according to the similarity conditions and the similarity between the target case and the historical cases.

[0090] Based on the above, the cultural-genetic algorithm disclosed in the embodiment of the present application uses a chromosome coding model such as formula (5) to represent an individual:

[0091]

[0092] wherein, sim

[0093] The average accuracy of prediction using each combination of weight values is used to evaluate the fitness of the individual.

[0094] The following is a specific description of the weight optimization method based on the cultural-genetic algorithm:

[0095]

[0096] After obtaining the optimal combination of weight values according to the above weight calculation method, the similarity is calculated according to the optimal combination of weight values, K best similar historical cases are retrieved as similar cases, and the GDQoS value of the target case is predicted according to the GDQoS value in the historical cases, and the calculation is as shown in formula (6):

[0097]

[0098] wherein, a characteristic attribute value of a target case, i.e., a service GQoS value to be predicted and a DQOS value, a GQoS value and a DQOS value of a retrieved historical case SC i SIM(TC0, SC i ) represents a similarity between the target case TC0 and the historical case SC i . The GQoS value and the DQOS value of the K best historical cases retrieved according to the similarity are substituted into equation (6) respectively, so as to obtain the predicted values of the GQoS and the DQOS of the component service in the current context.

[0099] S2, compare the general service quality and the domain service quality of the original component service with the general service quality constraint and the domain service quality constraint corresponding to the original component service respectively, and determine the failed original component service in the composite service according to the comparison result.

[0100] The general service quality constraint (hereinafter referred to as GQoS constraint) and the domain service quality constraint (hereinafter referred to as DQoS constraint) of the original component service can be determined based on the GQoS constraint and the DQoS constraint of the composite service.

[0101] In one embodiment, the GQoS constraint of the original component service is obtained by using a preference-aware global GQoS constraint optimization decomposition method. The specific obtaining method comprises:

[0102] obtaining the general service quality constraint corresponding to the composite service;

[0103] decomposing the general service quality constraint corresponding to the composite service into the general service quality constraint corresponding to each original component service in the composite service by using a social learning optimization algorithm.

[0104] When a user submits a composite service use request, the user often proposes a global GQoS constraint for the entire composite service. In order to obtain the local GQoS constraint that each component service needs to meet, it is necessary to optimize and decompose the global GQoS constraint proposed by the user into the local GQoS constraint that each component service needs to meet, and the local GQoS constraint is also the basis for judging the GQoS failure of the component service. However, the existing global QoS constraint optimization decomposition method usually takes the QoS value interval of all historical component services corresponding to an abstract component service as the GQoS value interval of the abstract component service, and performs global GQoS constraint optimization decomposition based on these value intervals. This processing method does not consider the GQoS preferences of different user groups, resulting in a too rough GQoS value interval and a low accuracy.

[0105] In order to improve the accuracy of global GQoS constraint optimization decomposition, the embodiment of the application proposes a preference-aware global GQoS constraint optimization decomposition method, and the main idea is as follows: first, according to the service use preference of the user group, the GQoS attribute value interval of each abstract component service is determined; then, the GQoS value interval corresponding to each abstract component service is divided into a set of discrete quality values (referred to as quality scales), and an improved SLO algorithm is used to find a set of optimal quality scale combinations for each abstract component service, and the quality scale combination is the local GQoS constraint that the abstract component service needs to meet.

[0106] In the swarm intelligence algorithm, a feasible solution is referred to as an individual, and all feasible solutions are referred to as a population. Specifically, the individual refers to a solution of the global GQoS constraint optimization decomposition problem; the population refers to all feasible solutions when the improved social learning optimization algorithm is used to solve the global GQoS constraint optimization decomposition problem. Different user groups have different local GQoS constraint preferences for the same component service of the combined service. In order to obtain the local GQoS constraint corresponding to the user preference, when performing global GQoS constraint optimization decomposition, the GQoS attribute value interval corresponding to each abstract service in the service combination process needs to be determined according to the GQoS preference of different user groups. When determining the historical GQoS value interval of the component service, the GQoS preference of the user group is considered, and the GQoS attribute value interval of each abstract component service is determined according to the historical record of the use of the combined service by different user groups. The specific operation process is as shown in Figure 2 First, the historical combined services with the same functional attributes are selected from the historical combined service set; then, according to the preferences of the users who have used these historical combined services, the users are classified to obtain multiple user groups; finally, according to the user classification result, the historical GQoS values of each type of abstract component service in the combined services with the same functional attributes in different user groups are counted, and the GQoS value interval perceived by the user preference is determined.

[0107] For each abstract component service in the service combination process, the value interval of each GQoS attribute is determined according to the GQoS value interval perceived by the preference, and each value interval is divided into a set of discrete quality values, and each discrete quality value is referred to as a quality scale of the attribute. Let the abstract component service S j There are k GQoS attributes, and the value interval of the i-th GQoS attribute is The division of the quality scale of the i-th GQoS attribute is shown in formula (7).

[0108]

[0109] wherein, and respectively represent the maximum value and the minimum value of the i-th QoS attribute of the abstract component service S j , d represents the number of discrete values of each quality level, represent the d-th discrete value of the i-th quality level of the abstract component service S j . By dividing the value interval of the k GQoS attributes in the abstract component service S j , k quality level sets are obtained. For each abstract component service, the quality level generation process is shown in Figure 3 . Wherein, Figure 3 the left side of which represents the historical values of the k GQoS attributes of the abstract component service S j ; Figure 3 the right side of which represents the quality level corresponding to each GQoS attribute obtained by dividing the value interval of each GQoS attribute.

[0110] In practical applications, the composite service usually needs to meet several global GQoS attribute constraints (supposed to be k) proposed by the user, therefore, when performing global GQoS constraint decomposition, each abstract component service needs to meet k local GQoS constraints. For each abstract component service, a quality level is selected for each GQoS attribute, and the selected quality levels are put together, and a quality level combination of the abstract service is obtained.

[0111] The goal of global GQoS constraint decomposition is to find a set of quality level combinations (QLC) for each abstract component service of the composite service process, so that when each abstract component service meets the set of quality level combinations, the aggregated GQoS constraint of all abstract component services can also meet the global GQoS constraint. At the same time, the number of candidate services under these quality level combinations is as large as possible, and the GQoS of the candidate services is as good as possible. Therefore, the number of candidate services meeting the quality level combination and the utility value of the quality level combination are used as the main indexes for evaluating the quality level combination. The evaluation formula of the quality level combination is shown in equation (8):

[0112]

[0113] Wherein, represents the i-th quality level combination of the abstract component service S j , represents the number of services meeting the i-th quality level combination, represents the utility value of the i-th quality level combination of the abstract component service S j . The abstract component service S jThe i-th quality scale combination (QLC) utility function of S is shown as formula (9).

[0114]

[0115] wherein, denotes the abstract component service S j The r-th QoS attribute value of the i-th quality scale combination, and denotes the abstract component service S j The maximum and minimum values of the r-th QoS of S r denotes the user preference, w r ∈(0, 1]. The goal of the global QoS constraint optimization decomposition is to find the best quality scale combination for each abstract component service, and the objective function of the global QoS constraint optimization decomposition problem is shown as formula (10).

[0116]

[0117] wherein, n is the number of abstract services in the composite service process, denotes the abstract service class S j The evaluation value of a quality scale combination.

[0118] In an embodiment, the essence of the global GQoS constraint optimization decomposition is a combination problem, and the embodiment of the application improves the social learning optimization algorithm according to the characteristics of the problem, so that the global GQoS constraint optimization decomposition problem can be solved efficiently. The social learning optimization algorithm is used to decompose the general service quality constraint corresponding to the composite service into the general service quality constraint corresponding to each original component service in the composite service, including:

[0119] The decomposition task of decomposing the general service quality constraint corresponding to the composite service into the general service quality constraint corresponding to each original component service in the composite service is mapped into an original individual;

[0120] The original individual is subjected to evolutionary operation and learning operation in the micro-space and the learning space respectively to obtain an evolved individual and a learned individual; the evolutionary operation includes crossover operation and / or mutation operation, and the learning operation includes observation learning and / or imitation learning;

[0121] The fitness values of the original individual, the evolved individual and the learned individual are calculated, and the original individual is subjected to greedy operation with the evolved individual and the learned individual respectively according to the fitness values, and the individual with a higher fitness value is retained;

[0122] The individual with a higher fitness value in the population is uploaded to the belief space according to the update frequency;

[0123] Replace the individual with poor fitness value in micro space with the individual in belief space.

[0124] The above-mentioned social learning optimization algorithm (SLO) is a kind of swarm intelligence algorithm simulating the evolution process of human social intelligence. The algorithm has three layers of collaborative evolution space, which are micro space, learning space and belief space. The micro space simulates the genetic variation process of human social intelligence evolution and provides individuals with excellent genes for the learning space. The individuals in the learning space improve their intelligence level through observation and imitation learning. The belief space extracts excellent individuals from the learning space as knowledge and stores them in the space. With the increase of iteration times, the knowledge is accumulated and deposited to form culture. The culture guides the evolution of the group in the micro space through the influence operation. The SLO algorithm has good optimization mechanism and collaborative evolution mechanism and has good optimization solving performance. However, the existing SLO algorithm cannot be directly used to solve the global QoS constraint optimization decomposition problem. Therefore, the embodiment of the application improves the SLO algorithm and designs a new learning operator in combination with the characteristics of the global QoS constraint optimization decomposition problem. The following introduces the individual coding format and the optimization operation in each evolution space:

[0125] (1) Coding mode of solution

[0126] The global GQoS constraint optimization decomposition aims to find the best quality level combination (QLC) for each abstract component service. Each abstract component service corresponds to a QLC, and a combination service contains multiple abstract component services. Therefore, the solution of the global GQoS constraint optimization decomposition problem is composed of a group of QLCs. The embodiment of the application uses a two-dimensional coding model to represent the individual X of the global GQoS constraint optimization decomposition problem i , and the structure is as shown in Figure 4 . Figure 4 In the formula, S1, S2, …, S n represent n abstract component services in the combination service, q1, q2, …, q k represent k GQoS attributes in the global GQoS constraint, S nk L e represents the e th quality level of the k th QoS attribute of the abstract component service S n . n1 L r , S n2 L u , S n3 L v , …, S nk L e represent the n th abstract component service S nA set of quality scale combinations (QLC). In this embodiment of the invention, formula (10) is applied to calculate the fitness value of an individual in the population, and the fitness value of an individual changes with the operations within the three-layer co-evolutionary space.

[0127] (2) Operations within the microspace

[0128] Genetic variation and evolution processes are performed within a microspace, primarily involving selection, crossover, mutation, and greedy operations. Selection is performed using a roulette wheel. The definitions of crossover and mutation operations are as follows:

[0129] (a) Crossover operation

[0130] Let X i and X j Let i be any two individuals, and i ≠ j. This embodiment of the invention uses a single-point crossover mode, where r is a random number in (0,1), and p... c Let r be the crossover rate, and d be a random integer in the range [1, k], representing the crossover position when two individuals cross over. If r < p c Then perform a crossover operation, setting X... i The first d-1 columns and X j Columns d to k are merged and treated as individual X. i New individuals after crossover Similarly, X i The d to k columns and X j The first d-1 columns are merged and treated as individual X. j New individuals after crossover Cross operations such as Figure 5 As shown in the figure This indicates that two individuals are performing an interleaved operation.

[0131] The fitness values ​​of the new individuals generated before and after crossover are calculated according to formula (10) (the new individuals are evolved individuals). Then, a greedy operation is performed to compare individual X. i and X j and The evaluation value is used to retain individuals with higher fitness values.

[0132] (b) Mutation operation

[0133] This invention employs a single-point mutation operation on individuals, specifically by randomly selecting individual X. i Given a set of quality scales, randomly select a new quality scale from the corresponding set of quality scales to replace the selected quality scale. The mutation operation is as follows: Figure 6 As shown. Figure 6 middle, For the new individual generated after mutation, m is a random integer in [1, k], representing individual X. iThe position where the mutation occurs. After the mutation operation is performed, the individual with higher fitness value is reserved between the old individual (the original individual) and the new individual (the evolved individual) according to the greedy operation.

[0134] (3) Operation in learning space

[0135] In the global GQoS constraint optimization decomposition problem, in order to ensure that the individual after participating in learning still satisfies the global GQoS constraint, the embodiment of the present application proposes a vertical learning mode, and the optimal part is learned from the excellent individual. When the target individual X i learns from the excellent individual X j , the evaluation value of each column quality scale in the individual X j is calculated, the column with the highest evaluation value in the individual X j is replaced with the corresponding column of the target individual X i , and the purpose of learning from the excellent individual is achieved. For positive QoS attributes such as availability and reliability, the evaluation function is as shown in formula (11). For reverse QoS attributes such as cost and time, the evaluation function is as shown in formula (12):

[0136]

[0137]

[0138] Wherein, GC k represents the global constraint of the kth GQoS attribute that the combined service needs to satisfy, n represents the number of abstract component services in the combined service, represents the kth GQoS attribute value of the lth abstract component service in the individual X i .

[0139] (a) Observational learning

[0140] The embodiment of the present application designs the observational learning operation as follows: the target individual only learns the optimal part (i.e. the optimal column of the excellent individual) of the excellent individual, and the other parts of the target individual are reserved. Let X i =(QLC i1 ,QLC i2 ,…,QLC ik ) be any individual in the population, be the current optimal solution, be the new individual generated after observational learning. The definition of the observational learning operation is as follows:

[0141]

[0142] Wherein, represents that the individual X i learns the current optimal individual X Gnew individual; X i (keep) represents the individual X i the part reserved after observation learning; X G (lopti) represents the individual X i to the optimal individual X G the part of learning; represents using the optimal individual X G the optimal column of replacement individual X i the corresponding column in. Figure 7 exhibits the individual X i the process of observation learning.

[0143] (b) imitation learning

[0144] The embodiment of the present application randomly selects two superior individuals with higher fitness values than the target individual from the population, and forms a learning group with the current individual to perform imitation learning, and the target individual learns the optimal part of the superior individual. The definition of imitation learning is as follows:

[0145]

[0146] wherein, represents the individual X i through the superior individual X r1 and X r2 new individual; X i (keep) represents the individual X i the part reserved after imitation learning; X r1 (lopti) and X r2 (lopti) respectively represent the individual X i to the superior individual X r1 and X r2 the part of imitation learning; represents using the superior individual X r1 and X r2 the optimal column of replacement individual X i the corresponding column in. Figure 8 exhibits the individual X i the process of imitation learning.

[0147] After the observation learning and imitation learning operations are performed, the fitness values of the original individual and the new individual (the new individual is the individual after learning) are calculated and compared, and the greedy operation is performed to retain the individual with a higher fitness value.

[0148] (4) Operation in belief space

[0149] There are two main operations in the belief space, which are the knowledge update operation (Update()) and the cultural influence operation (Influence()). The update operation in the belief space is as follows: first, the individuals in the learning space are arranged in descending order according to the fitness values; then, the top α individuals with high fitness values are extracted into the belief space to replace the original α individuals in the belief space, so as to realize the update and accumulation of knowledge. The update operation in the belief space is shown in formula (15):

[0150] α = p·β (15)

[0151] wherein α represents the number of individuals in the belief space, p represents the population size, and β represents the proportion of the number of excellent individuals in the belief space to the population size. The influence operation is to replace the individuals with poor fitness values in the micro space with the knowledge in the belief space, so as to guide the evolution of the population in a better direction and improve the convergence speed of the algorithm. When the algorithm runs to the iteration number that needs to be updated, the α individuals in the belief space are replaced with the α individuals with poor fitness values in the micro space; if the update interval is not reached, the influence operation is not performed. The influence operation in the belief space is shown in formula (16):

[0152]

[0153] wherein S represents an individual in the belief space, j ∈ [1, α]; represents an individual with poor fitness value in the micro space, i ∈ [1, α]; t is the current iteration number; and ε is an update interval parameter, which is equal to 5 in the embodiment of the present application, that is, the influence operation is performed once every five iterations.

[0154] The global GQoS constraint decomposition algorithm based on the improved social learning algorithm and preference perception is as follows:

[0155]

[0156] In one embodiment, the method for obtaining the domain service quality constraint corresponding to the original component service comprises: obtaining the domain service quality constraint that needs to be satisfied by the component service by using a statistical method.

[0157] It is assumed that S j (1≤j≤n) is an abstract component service in a composite service process, the domain service quality (DQoS) indicators of S t are <dq1, dq2,..., dqn>; and the expected value of the ith DQoS indicator of S j for each user in a class of users (including U users) is According to the statistical method, the expected value of the ith DQoS indicator of S jthe expectation of the i-th domain quality of service indicator of the original component service is Similarly, the expectation of S j The expectation of other DQoS indicators, i.e., the expectation of the DQoS indicators of the abstract component service S j The DQoS constraint to be met is also the DQoS constraint to be met by the j-th component service in the combined service.

[0158] S3, obtaining a corresponding replacement component service for the failed original component service.

[0159] The above, the determination of the replacement component service can be obtained by the clustering method, and the commonly used clustering method in the prior art includes a plurality of, for example: K-means algorithm of maximum minimum distance and weighted density (KWDM), minimum spanning tree heuristic segmentation method (DMSC) and the like.

[0160] When the candidate component service is multiple, the replacement component service is determined in the candidate component service set according to the GQoS and DQoS of the candidate component service. The acquisition method of the GQoS and DQoS of the candidate component service can be obtained according to the prediction method of the GQoS and DQoS of the original component service in step S1, which will not be repeated here.

[0161] In one embodiment, step S3 comprises:

[0162] Based on the minimum spanning tree heuristic segmentation clustering method, at least one candidate component service is determined in the component service library for the failed original component service, and the component service library includes a plurality of resource component services.

[0163] The above, specifically includes: taking the average similarity of the resource component service as the truncation distance, calculating the local density of each resource component service;

[0164] According to the size of the service local density, the resource component services are sorted, the points corresponding to the resource component services meeting the sorting condition are taken as the initial class cluster centers, and the remaining resource component services are distributed according to the similarity;

[0165] Mapping each class cluster to a node, mapping the similarity between class clusters to the weight of the edge between nodes, calculating the intra-cluster similarity and inter-cluster similarity of each class cluster, and using the kruskal algorithm to select the edge with the maximum weight to generate a minimum spanning tree;

[0166] Segmenting the edge with the minimum weight in the minimum spanning tree to obtain new class clusters, and storing the result as a new clustering result;

[0167] The segmenting operation is stopped when all edges are cut, the ratio of the intra-cluster similarity to the inter-cluster similarity of each clustering result is calculated, the ratio with the minimum value is selected as the final clustering result, and the candidate component service is obtained by fine tuning according to the K-Means algorithm.

[0168] In the foregoing, the embodiment of the application provides a service clustering method based on heuristic segmented clustering for a minimum spanning tree (Heuristic Segmented for MST Clustering based on Service Density, DMSC). The method can effectively cluster large-scale services on a network according to various attributes of the services, and lays an important foundation for discovery of the best candidate service.

[0169] Currently, the K-means algorithm based on maximum-minimum distance and weighted density (KWDM) is an improved K-Means clustering algorithm, which has good clustering performance, selects sample points with greater density as clustering centers in sequence, and determines the K value according to the ratio of the inter-cluster similarity to the intra-cluster similarity, thereby overcoming the randomness of selection of the clustering centers in the K-means algorithm and avoiding the problem that the clustering result falls into a local optimum, but the algorithm still has some limitations.

[0170] (1) In the selection of the K value, the KWDM algorithm stipulates that the number of clustering centers K can only be between 1 and n, and cannot well cope with the case that services are distributed relatively discretely and there are too many clusters.

[0171] (2) In the definition of the local density, the case that the number of clustering objects is the same but the tightness is different is not considered, which leads to insufficient accuracy of selection of the initial clustering centers and has a bad influence on the accuracy of the clustering result.

[0172] In view of the foregoing deficiencies, the embodiment of the application makes the following improvements to the KWDM algorithm:

[0173] (1) In view of the deficiency that the KWDM algorithm has inaccurate clustering results when clustering objects are distributed relatively discretely, the embodiment of the application takes the first several points with greater local density as the initial cluster centers, and determines the cluster with fewer clustering objects as an outlier, so as to improve the accuracy of clustering.

[0174] (2) In view of the case that the number of clustering objects is the same but the tightness is different within the same radius, a new local density function is designed, the similarity between services sim(s p ,s q ) is taken as part of the calculation of the local density of the services, the accuracy of the local density is improved, and the selection of the initial cluster centers is facilitated. The local density function is shown in formula (17), wherein s i ​​​denotes the number of services within the radius, denotes the circle denotes the number of service pairs within the circle.

[0175]

[0176] The embodiment of the application stores the initial clustering result in the form of a minimum spanning tree and processes the aggregation problem between the class clusters, which not only can preserve the initial clustering result, but also reduces the complexity of service clustering because the minimum spanning tree is a sparse graph. Based on the above-mentioned improvement, a heuristic segmented clustering algorithm based on the minimum spanning tree and service density (DMSC) is proposed, and the DMSC algorithm is applied to cluster services to obtain K services most similar to the failed service. The service clustering process based on DMSC is as follows:

[0177]

[0178] In Step 1, the average similarity of services is first calculated i ) as the truncation distance, and the local density density(s The services with the top local density values are added to the initial clustering center set CS, and the remaining services are distributed according to the maximum-minimum distance criterion; in Step 3, the intra-cluster similarity ICS(Ci) and the inter-cluster similarity ECS(Ci,Cj) of each class cluster are calculated, and the edge with the maximum weight W is selected to generate the minimum spanning tree MST by using the Kruskal algorithm. Wherein, the node N represents each class cluster, and ECS(Ci,Cj) represents the weight W of the edge; in Step 4, the edge with the minimum weight W in the MST is segmented (i.e., the edge with the minimum ECS(Ci,Cj) is segmented), and two new class clusters are obtained, which are taken as the new clustering result. The step is executed in a loop until all edges are cut off, and the segmentation operation is stopped; finally, by comparing the ratio of ECS(π) to ICS(π), the clustering result with the minimum ratio is selected, and the K-Means algorithm is used to fine-tune the result as the service clustering result π={C1,C2,C3,…,Cm}.

[0179] S4, replacing the failed original component service with the replacement component service to generate a reconstructed composite service.

[0180] The reconstructed composite service generated above can include part of the original component service, part of the replacement component service, or all replacement component services, but regardless of the above cases, the existing component service in the reconstructed composite service is valid.

[0181] Based on the combination service reconstruction method disclosed in the embodiments of the application, the embodiments of the application also provide a combination service reconstruction device, as shown in the accompanying drawings. Figure 9 The combination service reconstruction device comprises:

[0182] A service quality prediction module 901 is configured to predict the general service quality and the domain service quality of the original component service in the combination service.

[0183] A failed service determination module 902 is configured to compare the general service quality and the domain service quality of the original component service with the general service quality constraint and the domain service quality constraint corresponding to the original component service, respectively, and determine the failed original component service in the combination service according to the comparison result.

[0184] A replacement service determination module 903 is configured to obtain a corresponding replacement component service for the failed original component service.

[0185] A combination service reconstruction module 904 is configured to replace the failed original component service with the replacement component service to generate a reconstructed combination service.

[0186] In one embodiment, the service quality prediction module 901 is specifically configured to jointly predict the general service quality and the domain service quality of the original component service in the combination service by using a case-based reasoning method.

[0187] Further, the service quality prediction module 901 comprises:

[0188] A case representation module is configured to represent the original component service as a target case, wherein the condition feature of the target case comprises the context information of the original component service at the time of imminent running, and the decision feature of the target case comprises the general service quality and the domain service quality corresponding to the future running of the original component service.

[0189] A similar case determination module is configured to determine the similar case of the target case in a historical case library according to the similarity between the target case and the historical case.

[0190] A prediction module is configured to predict the general service quality and the domain service quality of the target case according to the general service quality and the domain service quality in the similar case.

[0191] In one embodiment, the similar case determination module is specifically configured to:

[0192] determine the optimal weight combination corresponding to different decision features by using a cultural-genetic algorithm;

[0193] calculate the similarity between the target case and the historical case according to the optimal weight combination;

[0194] According to the similarity condition and the similarity between the target case and the historical cases, similar cases are determined in the historical case library.

[0195] In one embodiment, the apparatus disclosed by the present application further comprises a general service quality constraint obtaining module, comprising:

[0196] A global general service quality constraint obtaining module is configured to obtain the general service quality constraint corresponding to the combined service.

[0197] A decomposition module is configured to decompose the general service quality constraint corresponding to the combined service into the general service quality constraints corresponding to each original component service in the combined service by using a social learning optimization algorithm.

[0198] In one embodiment, the decomposition module is specifically configured to:

[0199] The decomposition task of decomposing the general service quality constraint corresponding to the combined service into the general service quality constraints corresponding to each original component service in the combined service is mapped to an original individual.

[0200] The original individual is subjected to evolutionary operation and learning operation in the micro-space and the learning space respectively to obtain an evolved individual and a learned individual; the evolutionary operation comprises crossover operation and / or mutation operation, and the learning operation comprises observation learning and / or imitation learning.

[0201] The fitness values of the original individual, the evolved individual and the learned individual are calculated, and the original individual is subjected to greedy operation with the evolved individual and the learned individual respectively according to the fitness values, and the individual with a higher fitness value is retained.

[0202] The individual with a better fitness value in the population is replaced by a better individual in the belief space.

[0203] In one embodiment, the individual to be updated in the micro-space is replaced by the updated individual in the belief space, comprising:

[0204] The updated individual is uploaded to the belief space according to an update frequency.

[0205] The individual to be updated in the micro-space is replaced by the updated individual in the belief space.

[0206] In one embodiment, the apparatus disclosed by the present application further comprises:

[0207] A domain service quality constraint obtaining module is configured to obtain the domain service quality constraint corresponding to the original component service, comprising:

[0208] The domain service quality constraint required to be satisfied by the component service is obtained by using a statistical method.

[0209] In one embodiment, the replacement service determining module 903 is specifically configured to:

[0210] Determine at least one candidate component service for the failed original component service in the component service library based on a minimum spanning tree heuristic segmentation clustering method, the component service library comprising a plurality of resource component services.

[0211] Further, the replacement service determination module 903 is specifically configured to:

[0212] Determine at least one candidate component service for the failed original component service in the component service library based on a minimum spanning tree heuristic segmentation clustering method, the component service library comprising a plurality of resource component services.

[0213] The above specifically comprises: taking the average similarity of the resource component services as a truncation distance, and calculating the local density of each resource component service;

[0214] According to the size of the service local density, the resource component services are sorted, the points corresponding to the resource component services satisfying the sorting condition are taken as initial cluster centers, and the remaining resource component services are distributed according to the similarity;

[0215] The clusters are mapped into nodes, the similarity between the clusters is mapped into the weight of the edges between the nodes, the intra-cluster similarity and the inter-cluster similarity of each cluster are calculated, and the Kruskal algorithm is used to select the edge with the maximum weight to generate a minimum spanning tree;

[0216] The edge with the minimum weight in the minimum spanning tree is segmented to obtain new clusters, and the result is stored as a new clustering result;

[0217] When all the edges are cut, the segmentation operation is stopped, the ratio of the intra-cluster similarity to the inter-cluster similarity of each clustering result is calculated, the ratio with the minimum value is selected as the final clustering result, and the candidate component service is obtained.

[0218] In one embodiment, the replacement service determination module 903 is further configured to, when the candidate component service is multiple, determine a replacement component service from the candidate component services according to the general service quality and the field service quality of the candidate component services.

[0219] In addition, the embodiment of the present application further provides an electronic device, comprising: one or more processors; and a memory associated with the one or more processors, the memory being used to store program instructions, the program instructions being read and executed by the one or more processors to perform the combination service reconstruction method disclosed in the above embodiment.

[0220] Wherein, Figure 10An exemplary system architecture of the electronic device is shown, which can specifically include a processor 1010, a video display adapter 1011, a disk drive 1012, an input / output interface 1013, a network interface 1014, and a memory 1020. The processor 1010, the video display adapter 1011, the disk drive 1012, the input / output interface 1013, the network interface 1014, and the memory 1020 can be communicatively connected through a communication bus 1030.

[0221] The processor 1010 can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, etc., and is configured to execute related programs to implement the technical solutions provided in the present application.

[0222] The memory 1020 can be implemented in the form of a ROM (Read Only Memory), a RAM (Random Access Memory), a static storage device, a dynamic storage device, etc. The memory 1020 can store an operating system 1021 for controlling the operation of the electronic device, a BIOS (Basic Input Output System) for controlling the low-level operation of the electronic device. In addition, a web browser 1023, a data storage management system 1024, and a device identification information processing system 1025, etc. can also be stored. The device identification information processing system 1025 can be an application program that specifically implements the above-mentioned steps in the embodiments of the present application. In summary, when the technical solutions provided in the present application are implemented by software or firmware, the related program codes are stored in the memory 1020 and executed by the processor 1010.

[0223] The input / output interface 1013 is configured to connect an input / output module to realize information input and output. The input / output module can be configured as a component in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. The input device can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output device can include a display, a speaker, a vibrator, an indicator light, etc.

[0224] The network interface 1014 is configured to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as USB, network cable, etc.) or a wireless manner (such as mobile network, WIFI, Bluetooth, etc.).

[0225] Bus 1030 includes a path for transferring information between the various components of the device, such as processor 1010, video display adapter 1011, disk drive 1012, input / output interface 1013, network interface 1014, and memory 1020.

[0226] In addition, the electronic device can also obtain the information of the specific obtaining condition from the virtual resource object obtaining condition information database, for condition judgment, etc.

[0227] It should be noted that although the above device only shows the processor 1010, video display adapter 1011, disk drive 1012, input / output interface 1013, network interface 1014, memory 1020, bus 1030, etc., but in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only contain the components necessary to implement the scheme of the present application, and does not have to contain all the components shown in the figure.

[0228] In particular, according to the embodiments of the present application, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, the embodiments of the present application include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for executing the method shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from the network through the communication device, or installed from the memory, or installed from the ROM. When the computer program is executed by the processor, the above-mentioned functions defined in the method of the embodiments of the present application are executed.

[0229] It should be noted that the computer readable medium in the embodiments of the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. In the embodiments of the present application, the computer readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus or device. In the embodiments of the present application, the computer readable signal medium can include a data signal carrying computer readable program code in a baseband or as part of a carrier wave. Such a propagated data signal can take any of a variety of forms, including but not limited to electro-magnetic, optical, or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium that is not a computer readable storage medium and that can be used to carry or store a program for use by or in connection with an instruction execution system, apparatus or device. The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to wire, cable, RF, etc., or any suitable combination thereof.

[0230] The computer readable medium described above can be contained in the server described above; or can exist separately and not be assembled into the server. The computer readable medium described above carries one or more programs, when the one or more programs are executed by the server, the server: in response to detecting that the peripheral mode of the terminal is not activated, acquires the frame rate of the application on the terminal; when the frame rate meets the off-screen condition, judges whether the user is acquiring the screen information of the terminal; in response to the result of the judgment that the user is not acquiring the screen information of the terminal, controls the screen to enter the immediate dim mode.

[0231] Computer program code for carrying out operations of the embodiments of the present application can be written in one or more programming languages or combinations of languages including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages such as "C" or the like. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0232] The above describes the technical solutions provided by the present application in detail, and the principles and implementation manners of the present application are described by applying specific examples in the embodiments of the present application. The above examples are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In conclusion, the content of the specification should not be understood as a limitation of the present application.

[0233] All the optional technical solutions described above can be combined to form optional embodiments of the present application, and will not be described one by one here.

[0234] The above describes the technical solutions provided by the present application in detail, and the principles and implementation manners of the present application are described by applying specific examples in the embodiments of the present application. The above examples are only used to help understand the method of the present application and its core idea; meanwhile, for those skilled in the art, the specific implementation manners and application ranges can be changed according to the idea of the present application. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A method for reconstructing composite services, characterized in that, include: Predict the general service quality and domain service quality of the original component services in the composite service; The general service quality and domain service quality of the original component service are compared with the general service quality constraints and domain service constraints corresponding to the original component service, respectively. Based on the comparison results, the original component service that fails in the composite service is determined. The method for obtaining the general service quality constraints corresponding to the original component service is as follows: determine the value range of the general service quality attributes of each abstract component service based on the service usage preferences of the user group; The general service quality range corresponding to each abstract component service is divided into a set of discrete quality values, called quality scales. Based on the improved social learning algorithm, an optimal combination of quality scales is found for each abstract component service. This combination of quality scales is the local general service quality constraint that the abstract component service needs to satisfy, thus obtaining the general service quality constraint corresponding to the original component service. In the operation of the learning space in the improved social learning algorithm: longitudinal learning is performed. When the target individual is learning from the excellent individuals, the quality scale of each column in the excellent individuals is evaluated, and the column with the highest evaluation value in the excellent individuals is replaced with the corresponding column of the target individual. Obtain the corresponding replacement component service for the failed original component service; Replace the failed original component service with a replacement component service to generate a refactored composite service.

2. The combined service reconstruction method as described in claim 1, characterized in that, Using a case-based reasoning approach, the general service quality and domain service quality of the original component services in the combined service are jointly predicted, including: The original component service is represented as a target instance. The conditional characteristics of the target instance include the context information of the original component service when it is about to run. The decision characteristics of the target instance include the general service quality and domain service quality corresponding to the original component service when it runs in the future. Based on the similarity between the target instance and historical instances, similar instances of the target instance are identified in the historical instance database. Predict the general service quality and domain service quality of the target instance based on the general service quality and domain service quality of the similar instances.

3. The combined service reconstruction method as described in claim 2, characterized in that, The step of determining similar instances of the target instance in the historical instance database based on the similarity between the target instance and historical instances includes: The optimal weight combination corresponding to different decision characteristics is determined using a culture-genetic algorithm. The similarity between the target case and the historical case is calculated based on the optimal weight combination. Based on similarity criteria and the similarity between the target case and the historical case, the similar case is determined in the historical case database.

4. The combined service reconstruction method as described in claim 1, characterized in that, The step of obtaining a corresponding replacement component service for the invalid original component service includes: Based on the minimum spanning tree heuristic segmented clustering method, at least one candidate component service is determined for the failed original component service in the component service library, which includes multiple resource component services; wherein, the candidate component service includes: The average similarity of the resource component services is used as the cutoff distance to calculate the local density of each resource component service. The resource component services are sorted according to the size of the local density of the services. The points corresponding to the resource component services that meet the sorting conditions are used as the initial cluster centers, and the remaining resource component services are allocated according to similarity. Each type of cluster is mapped to a node, and the similarity between clusters is mapped to the weight of the edge between nodes. The intra-cluster similarity and inter-cluster similarity of each type of cluster are calculated, and the Kruskal algorithm is used to select the edge with the largest weight to generate the minimum spanning tree. The edge with the smallest weight in the minimum spanning tree is segmented to obtain a new cluster, and the result is stored as a new clustering result. Stop the segmentation operation when all edges have been cut, calculate the ratio of intra-cluster similarity to inter-cluster similarity for each clustering result, select the one with the smallest ratio as the final clustering result, and obtain the candidate component service.

5. The combined service reconstruction method as described in claim 1, characterized in that, The method further includes: When there are multiple candidate component services, the replacement component service is determined from the set of candidate component services based on the general service quality and domain service quality of the candidate component services.

6. A composite service reconstruction apparatus for implementing the composite service reconstruction method of claim 1, characterized in that, include: The service quality prediction module is used to predict the general service quality and domain service quality of the original component services in the composite service. The failed service determination module is used to compare the general service quality and domain service quality of the original component service with the general service quality constraints and domain service constraints corresponding to the original component service, respectively, and determine the failed original component service in the combined service based on the comparison results. The replacement service determination module is used to obtain the corresponding replacement component service for the failed original component service; The composite service reconstruction module is used to replace the failed original component service with the replacement component service to generate a reconstructed composite service.

7. An electronic device, characterized in that, include: One or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the combined service refactoring method as described in any one of claims 1 to 5.

8. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the combined service reconstruction method as described in any one of claims 1 to 5.

Images