A New Simplified Decision Tree Process Based On Clustering Technique To Overcome Over Fitting Problem For Better Decision Making

Decision tree is a decision support tool that uses a tree-like graph or model of decisions and their possible consequences. Overfitting is one of the problems in decision tree cause by over-complex trees that do not generalize the data well which downgrade the performance of decision tree machine learning classification. Any fluctuation in real data might cause inaccuracy and poor predictive performance .Mobile classification application has emerged as the result of the advancement in mobile technology. It has been used as the tool for mobile healthcare, emergency management, mobile policing, mobile commerce and mobile banking. These mobile application prove to save time and increase productivity for those who need to make rapid decisions in real time. Overfitting problem might become one of the obstacles in developing high performance mobile classification/predictive application since accurate classification is the key of mobile classification application. Therefore, this research proposes to overcome the overfitting problem in decision tree by simplify the decision tree process using clustering method. Currently, clustering method has been used to improve the decision tree accuracy and robustness, and the results were significant. The research aims to identify the main factors that contribute to the simplification of a decision tree induction and how clustering technique can optimize the decision making of a decision tree. In order to evaluate the performance of the proposed method, a game will be developed on mobile device that incorporate the decision tree, then uses to detect dyslexic children. As the result of this novel idea a system that utilize the self-learning concept and exclude the phonological barrier of a language can be develop to screen dyslexic children through mobile game.