High dimensional document classification using novel similarity function

Abstract

Document dimensionality is a major concern and worrying factor when high dimensionality documents are used for classification. Reducing the dimensionality can have both positive and negative effects. If dimensionality reduction is not appropriate then the classification performed using the reduced dimensionality documents may not give good classification results. Our previous research was focused on addressing dimensionality reduction using novel similarity function, but it did not address text classification. This paper addresses the classification task performed by applying the proposed similarity function. Experiment results prove the classifier performance with dimensionality reduction is better to the performance without dimensionality reduction.