Decision Trees: A comparison of various algorithms for building Decision Trees

Decision Trees are a decision support tool that contains tree like graph of decisions and the possible consequences. They are commonly used in different real world scenarios ranging from operations research to classifying a specie in a phylum given its features. The Decision Tree is implemented using traditional ID3 algorithm as well as an evolutionary algorithm for learning decision trees in this paper. The Traditional Algorithm for learning decision trees is implemented using information gain as well as using gain ratio. Each variant is also modified to combat over-fitting using pruning. The Evolutionary Algorithm is implemented with fitness proportionate and rank based as their selection strategy. The algorithm is also implemented to have complete replacement and elitism as replacement strategy. The two algorithms are compared based on their accuracy, precision and recall by varying the aforementioned parameters on the datasets taken from UCI Machine Learning repository[2]. The time taken for learning the Decision Tree by each algorithm corresponding to each setting is also compared in this paper.