The Improvement Of Two Typical Classification Algorithms

Data mining extracts the useful information from a lot of data. Classification is one of the important functions of data mining and has been widely used in many fields, such as medical treatment, insurance, finance. The different classification methods have their advantages and disadvantages. The data's accuracy may be different by using the different classification methods for the same data.Bayesian algorithm is often used, because of its simple algorithm and its high accuracy. When the assumption of attribute independence does not hold, naive Bayesian algorithm possibly leads to misjudgment in types of the will-be-tested samples. When the will-be-tested samples have the same probabilities in all categories, it is unable to judge the type of samples. Three improved algorithms are proposed for the limitations of the above algorithm in this paper and experiments are made in the mushroom data. Experimental results show that the accuracy of the improved algorithm is much higher than the accuracy of naive Bayesian algorithm.Rough set is another important technology of classification. The attribute reduction is an important problem in rough set theory. It can maintain the classification of the knowledge base and decision-making on the same conditions, delete the irrelevant or unimportant attributes. It can receive different reduction results by using the different reduction algorithms in a given information system. And the accuracy of different attribute reduction are not the same accuracy, the classification accuracy of some attribute reduction result may be much lower than the classification accuracy of another reduction. In view of this situation, the algorithm based on attribute frequency and the lower approximation of the attribute reduction is proposed in this paper and compared with the two other attribute reduction algorithms. Experimental results show that the accuracy of the proposing algorithm for attribute reduction is much higher.