A generalized probabilistic approach for managing inconsistency to improve classifier accuracy

Removal of inconsistency from a data set contributes significantly in improving classification accuracy. Inconsistency occurs when attributes of objects have same value but they belong to different classes. Inconsistency is either inherent in the data set or appear during different data preprocessing steps, like discretization, dimensionality reduction and missing value prediction. The aim of the paper is to develop a generalized inconsistency handling scheme based on probability distribution unlike the previous methods which are context dependent. We propose two algorithms to remove inconsistency by assigning class labels to the objects afresh based on the statistical properties of the training data set. The ultimate goal of this research work is to generate consistent data which provide superior classification accuracy compare to the original data set. The proposed methods are verified with real life intrusion domain NSL-KDD data set for establishing our claim.