Comparison of Machine Learning Techniques for Prediction Problems

Abstract

Last three decades, artificial intelligence have become mandatory for human lives since the revolution of computers. As a part of it, Machine Learning and the proposed techniques are used to solve the real life problems instead of humans by using their extended capabilities for classification, optimization, and as well as prediction. However, there is a big mystery for the optimum technique for any kind of data and problem domain. In this paper, preliminary experiments are performed using recent and challenging nine datasets and four popular machine learning techniques in order to determine optimum technique for prediction domain. Obtained results are analyzed using three evaluation methods; MSE value, EV and R2 scores. Obtained results showed that, increment or degradation of instances in datasets do not affect the performance of the techniques directly and neural network algorithms produce higher and more steady prediction rates.