PROJECT TITLE :

Discretization Using Combination of Heuristics for High Accuracy with Huge Noise Reduction

ABSTRACT:

In the course of time, a number of discretization algorithms have been developed; however, the issue of how to discretize data in a way that is both accurate and efficient is still an open question. A brand new discretization algorithm called SPID5 is proposed in this piece of research. It is based on a combination of two heuristics, one of which is local and the other of which is global. Both of these heuristics are supervised, and their combination results in a significant synergy. The local heuristic is the well-known information gain of the continuous attributes, and the global heuristic is a novel concept of iteratively reducing noise in the data set. Both heuristics can be thought of as information gain. The noise can be reduced by successively lowering the pseudo deletion count of the data set that is to be discretized. This brings about the desired effect. The performance of the SPID5 algorithm is compared with that of three well-known and time-tested discretization algorithms. In addition, six state-of-the-art classifiers and 35 real-world data sets sourced from the standard UCI data repository are utilized in this comparison. Not only in terms of classification accuracy, but also in terms of noise reduction in the data sets, SPID5's performance compares favorably with that of all three existing discretization algorithms it is compared with. This is the case both in terms of the noise reduction and the classification accuracy.