Generative Method: A Statistical Based Discretization for Interval-Based Classification using Decision Tree Classifier

One of the renowned preprocessing technique is discretization, which plays a vital role in data transformation and gaining valuable knowledge from huge volume of data. The intervals obtained can significantly enhance the classification accuracy, reduce computational complexity, and enhance the robustness of machine learning models. A major consideration in data point representation is necessary before performing discretization to ensure an efficient analysis from the optimal bin boundaries. Thus, the purpose served from this study is to highlight the significance of distribution patterns of the data points in determining the optimal number of bins. A generative discretization method is proposed and compared with traditional widely used methods. Experimental evaluations demonstrate that our method achieves comparable accuracy by applying to a decision tree based model that enhance the flexibility for complex datasets by leveraging statistical measure for interval formation.