Classification and prediction by decision trees and neural networks

Abstract

In this paper we present comparative study of two frequently used methods for prediction and classification in data mining. These methods are decision trees and neural networks. Decision trees with J48 and ID3 algorithms are used to solve common classification problems where the data sets have several non-category attributes and one category attribute. In this case we need to predict category attribute which depends on the others. Neural networks have wide utilization including function approximation, data processing, prediction, classification etc. Technical neural networks offer profoundly different approach to solve problems in comparison with other methods. It could be very interesting to compare these dissimilar methods in terms of efficiency, execution speed and error rates. We demonstrate results of this comparison.