An Extended C4.5 Classification Algorithm using Mathematical Series

Abstract

The algorithm C4.5 is an efficient decision tree based classification, which is derived from the ID3 approach. C4.5 is also a rule based classification algorithm. The main importance of the C4.5 algorithm is that it can deal with categorical data, over fitting of data and handling of missing values. The performance of C4.5 is superior to ID3 even with equal number of attributes. The EC4.5 (Exponential C4.5) is an extension of C4.5 algorithm which uses exponential of split value to predict the gain of attributes and handled the set back reported in C4.5. However the EC4.5 has some misclassification of data and to avoid this problem a new technique is introduced. This paper proposes a proficient technique TMC4.5 (Taylor-Madhava C4.5) to reduce the uncertainty in classification of data by integrating an exponential split value in EC4.5 and sin splitting value derived from the Madhava series. By using this technique an optimized gain value is obtained that reduces uncertainty. From the obtained result the TMC4.5 has far better results than the C4.5 and EC4.5 algorithms.