Mutual Information Gain Feature Selection Technique based on Bayes Classifier for high dimensional text data classification

Abstract

The text classification is based on constructing a model through learning from training examples to automatically classify text documents. With size of text document repositories grows rapidly storage requirement and computational cost of model learning become higher. The instance selection solve the above issues by reducing data size by filtering out noisy data from given training dataset. The existing work presented a biological based genetic algorithm (BGA) for effective and efficient text classification. The BGA fits a biological evolution into evolutionary process most streamlined process complies with reasonable rules. After long term evolution organisms find the most efficient way to allocate resources and evolve. It requires least computational time and provides better classification accuracy than GA. This method did not provide any feature selection criteria and marginal classification accuracy. The dimensionality of text datasets is very high. The propose work presents an efficient mutual information gain feature selection (MIGFS)technique based on naive bayes classifier for high dimensional text data classification.