Classification algorithms based on fisher discriminant and perceptron neural network

In this paper, we exploit a new method of implementing mining classification, i.e., Fisher classification algorithm. In comparison with the decision-tree ID3 algorithm and its improved algorithm that is based on the criterion of choosing the split attributes according to information gain ratios and simple Bayes classification algorithm, we find that Fisher classification algorithm has a higher predictive accuracy and relatively less computation effort. Due to the sensitiveness of these methods mentioned above to noise, we propose a perceptron neural network classification algorithm, which has the stronger noise-rejection ability.