A Weighing Based Feature Selection and Monotonic Classification (WFSMC) for the Effective Prediction of Outliers

Advancement in computer technologies in data processing explores the capabilities of research, communication and services in recent years. Knowledge Discovery on Database (KDD) is an essential process on data processing. Data mining is one of the form of KDD. Many feature selection and classification algorithms are used to select the relevant features and classify them according to criteria in data mining applications. The enormous quantity of high dimensional data leads to challenge in traditional data mining techniques. The employment of classifiers in data has deteriorated the performance of machine learning process since it sensible to noise, relevant or irrelevant features. The ordering of decision values on the basis of relevant features improve the performance of the classification process. In this paper, Weighing based Feature Selection and Monotonic Classification (WFSMC)are proposed to know the importance of relevant feature and design the feature selection based on constraints. This paper discusses the new feature weighing classifier on the basis of imputation methods in two ways: 1) Estimation of new distribution values for each feature, 2) Statistical test to measure the changes between original and distributed values. Based on the principle of large margin, the new feature selection algorithm evolved by introducing the monotonic constraints. The performance analysis such as no. of attributes, accuracy, precision and recall shows that proposed weighing based feature selection yields better results than the traditional algorithms.