Ontology-based soft computing and machine learning model for efficient retrieval

Unstructured and unorganized data always degrade the performance of search techniques and produce irrelevant results in response to the query as well as decrease the speed of retrieval results. Ontology in semantic web (SW) provides an adequate solution to represent the knowledge, because of its backbone knowledge of an application or domain. But, domain ontology has three basic problems while retrieving useful knowledge from a domain ontology: (a) structuring/arrangement, (b) unnecessary knowledge reduction, selection and extraction, and (c) speeding up the retrieval process. To resolve these problems, we proposed multi-level k-mean clustering approach with rough set and Bayesian network model for ontology (MLK-rBO). The proposed model works in four different phases—clustering, knowledge discovery, building a probabilistic network, and model evaluation. The model ensembles three different techniques, namely clustering, rough set (RS), and Bayesian network (BN). Finally, the proposed model is tested with statistical parameters and compared with other models, namely decision tree (DT), random forest (RF), and support vector machine (SVM) to evaluate performance. By analyzing experimental results, we observed that the MLK-rBO gives better accuracy: 98.36% for survey data (fever) and 86% for Wine quality data than available models.