MULTITHREADED FUZZY LOGIC BASED WEB SERVICES MINING FRAMEWORK.PDF

Abstract

Multithreaded Fuzzy Logic based Web Services Mining Framework Finding valuable and attractive web services is becoming difficult, due to massive number of web services. Requirement of web services mining like data mining is vital these days. Comparative study of web services composition with mining concepts is presented in this report. A web services mining frame work, based on fuzzy logic, fuzzy set theory and fuzzy matching algorithm is proposed. This framework helps in finding valuable services and composing those services into composite web services. Mined services are further filtered in the rules matching and evaluation phases where specified rules are matched. Framework is tested with different UDDI registries of large sizes and the results are compared with existing techniques. The proposed model is divided into different steps and phases, to reduce the model complexity and simplify different integrating processes. The problems, faced in mining process are complexity of the search space and pattern matching. The complexity model is targeted by introducing the concept of threading for parallel processing. A new thread is initiated for every member of fuzzy set and mines the search space for required computation. This parallel processing approach helps in optimizing the search and matching process and for efficient discovery of individual web services and composition of web services. The first step in proposed framework is scope and rules specifications. Scope and the rules are specified by a web service domain expert and these are according to required mining results. For example, domain expert is looking for web services, related to traveling or in the field of medicine. Rules specified by the domain expert will be matched in constraint satisfaction and evaluation phases for filtering and validating of found web services and their compositions. Based on the scope, specified by the domain expert and weights, fuzzy set is generated and accordingly assigned to each number of fuzzy set. Weights are calculated based on the probability model and with the help of local database. This local database is used to store members of fuzzy set and helps in calculating weights. Every member of the fuzzy set is used as v input to the next searching phase and after the phase of assigning weights, a new thread is initiated for every member of the fuzzy set. Based on the fuzzy matching algorithm,this thread explores the UDDI registry and looks for relevant services. Outputs of the found web services are further used for discovery of web services, which are using these as their input parameters for composing individual web services into composite web services. Web service mining results sorted in the indexing phase based on weights assigned and these sorted results are filtered in the rules satisfaction phase, where constraint specified in the first phase are matched with the publisher’s constraint. Publisher specifies any service relevant constraint in the web service description document and at this step of our proposed model, these rules are satisfied for filtering and validation of found results. These filtered results are used as input to evaluation phase where these results are gone through objective and subjective evaluation. The performance of the proposed approach is evaluated using different factors like precision, recall and f-measure. Framework is tested for web services mining and the values for precision, recall and f-measure are calculated. Also, these values compared with the existing frameworks shown where proposed framework has improved the web services mining. After discovery, the services are available for composition. Mining time for UDDI registries of different sizes is recorded. At the end, comparison is given with an existing technique to present the improvements of proposed framework.