AN EFFECTIVE QUASI OPPOSITIONAL SOCIAL SPIDER OPTIMIZATION BASED FEATURE SELECTION WITH PATTERN MATCHING BASED CLASSIFICATION MODEL

Classification is a process of properly determining the targeted class for an unlabeled data instance by learning from the instances defined by a collection of features and a class label. The instance based classification model has gained significant attention because of its straightforwardness and effective outcome. But several models are highly sensitive to noise and become inappropriate for real time problems. This paper introduces an efficient Quasi Oppositional Social Spider Optimization (QOSSO) based Feature Selection (FS) with Pattern Matching based Classification (PMBC) model, named as QOSSO-PMBC. The presented model involves two major processes namely feature selection and classification. Primarily, the QOSSO based FS process is executed to determine an optimal set of features. In addition, the quasi oppositional based learning (QOBL) concept is introduced to the classical SSO algorithm with an intention of enhancing the convergence rate. Besides, the PMBC model is employed to identify the proper patterns and determine the class labels of the feature reduced dataset. The experimental results of the QOSSO-PMBC model are validated against two benchmark dataset namely contact lenses and diabetes dataset. The experimental values denoted that the QOSSO-PMBC model has resulted in a maximum accuracy of 91.61% and 95.34% on the applied contact lenses and diabetes dataset.