Optimization of C5.0 classifier using Bayesian theory

Abstract

C5.0 classifier is optimized using Bayesian theory. The C5.0 algorithm is a classifier that discovers patterns in data and uses them to make accurate predictions. It classifies data objects based on the information gain of its attributes. Though it responds to noisy and missing data, its accuracy can be improved upon. This research work proposes a post pruning decision tree algorithm that will use C5.0 as its base and Bayesian posterior theory as an enhancer. Post pruning is performed by evaluating the decision tree using Bayesian posterior theory. Bayesian theory uses probability to judge the relative validity of hypothesis in terms of noisy and uncertain data. The proposed algorithm attempts to support low memory usage, higher accuracy and improved speed with the help of smaller decision trees. It will also reduce the risks associated with over-fitting.