Fast Numerical Threshold Search Algorithm for C4.5

Abstract

This paper presents a new algorithm to improve the speed of threshold searching process in C4.5 by using the technique of genetic algorithms. In the threshold searching process in C4.5, the values in a numerical attribute are sorted first and then the mid-point between every two consecutive values is calculated and designated as a candidate threshold. This process can be time consuming and it is not practical for large data. Our algorithm generates a population of possible thresholds and converges to the best threshold value rapidly. Our experimental results have shown that significant time reduction has been achieved by using our algorithm in threshold searching process.