A meta-heuristic approach for developing PROAFTN with decision tree

Machine learning algorithms known for their performance in using historical data and examples to predict and classify unknown instances. Decision tree is an efficient machine learning approach that can use data only without the involvement of decision maker to improve the decision making process. Multi-Criteria Decision Analysis (MCDA)is another paradigm used for data classification. In this paper, we propose a new fuzzy classification method based on MCDA called PROAFTN. To use PROAFTN, a set of parameters need to be established from data. The proposed approach uses data pre-processing and canonical genetic algorithm (GA) for obtaining these parameters from data. The generated models have been applied on popular data selected from several application domain, health, economy, etc. According to our experimental study, the new model performs significantly better than decision trees according in terms of accuracy and the interpretation of the decision rules.