A comparative study of Naive Bayes Classifiers with improved technique on Text Classification

Experiment was carried out on imbalanced data having positive and negative labels as 0 and 1. These datasets after training were tested on Gaussian Naive Bayes, Bernoulli Naive Bayes and Multinomial Naive Bayes with improved technique using tf-idf and ngram. The results obtained were then compared with old model result that make use of BagofWords. On testing it is found that there is a 2-3% improvement in the model's performance.