Lazy learning associative classification with hybrid feature selection

Abstract

Lazy learning associative classification is one of the associative classification methods in which it delays the generalization of the training data until it receives a test query. It lacks in performance due to availability of many features in the dataset. All the features do not contribute classification system. It is important to choose the most appropriate features to identify the class of unseen test tuples. This paper shows how hybrid feature selection method can be applied to lazy learning associative classification to overcome this issue. The proposed method integrates a forward selection and backward elimination approach of feature selection methods that leads to good selection of attributes and better accuracy. Experimental results of the proposed system are visibly positive in comparison to the traditional and existing associative classification methods.