Iterative Threshold-Based Naive Bayes Classifier: Further Interpretability with p-Values

Sentiment Analysis is a collection of techniques adopted to classify Natural Language texts into a set of sentiments (i.e., positive-neutral-negative), but processing and understanding human language is a challenging task. Forby, Machine Learning algorithms are becoming more and more complicated as well, specializing to maximize their performances among over the interpretability. With this paper, we consider the recently proposed explainable classifiers (Threshold-based Naive Bayes and iterative Threshold-based Naive Bayes) and the Central Limit Theorem to extend them to classical statistical tests. In view of that, not only has their interpretability been improved, but their implementations will be more stable, providing more results consistency.