The impact of fuzzy discretization�s output on classification accuracy of random forest classifier

Random Forest is known as among the widely used classification algorithms by researchers and machine learning enthusiast in solving classification problems. Recently, fuzzy discretization has been paired with Random Forest (RF) classifier to enhance the classification accuracy of Random Forest classifier when dealing with continuous variables. However, there are many different opinions on whether there is a need to perform discretization in data pre-processing for tree-based classifiers such as J48, Decision Tree and Random Forest. On top of that, it is known that different classification algorithms produce different classification accuracies depending on the type of data used. In other words, the output of data discretization process. Thus, to unravel this mentioned hypothesis, this study intends to shed some lights on the impact of different fuzzy discretization�s output on the classification accuracy of Random Forest classifier. In this study, three version of simulations were done with different fuzzy discretization output. Those fuzzy discretization�s outputs are 1) without fuzzy discretization 2) with fully fuzzy discretization and 3) with partial fuzzy discretization. Then, classification phase is done through Random Forest classifier and the classification accuracy for all the simulation versions were observed, recorded, and analyzed. © 2020, World Academy of Research in Science and Engineering. All rights reserved.