Using Neural Networks for Classification Tasks – Some Experiments on Datasets and Practical Advice

Abstract

This paper gives a general introduction to neural networks, and then provides a more detailed description of one type of neural network, namely the multi-layer perceptron. Four experiments on datasets are then reported. These compare the performance of a multi-layer perceptron with induced decision trees, and show some of the strengths of neural networks. A further dataset is then used to demonstrate some of the problems which may occur if there are insufficient data. The paper concludes with some general practical advice.