ANALYSIS OF OPTIMIZATION TECHNIQUES IN CHI- SQUARED AUTOMATIC INTERACTION DETECTION
G. M. Kadhar Nawaz · 2014
In the context of pattern classification, one of the major issues discussed by most of researchers is ‘curse of dimensionality’ problem which occurs in data classification because the data processed in most of the application is high-dimensional feature space.So, the essential consideration here is that irrelevant features should be identified which causes less classification accuracy and the main motto is to find a minimum set of attributes from the initial set of data helping to make the patterns easier to understand along with improved classification accuracy and reduced learning time. Therefore, the selection of feature set is the process to search for an optimal feature subset from the initial data set without compromising the classification performance and efficiency in generating classification model. In this paper, we develop a hybrid classifier by combining CHAID and genetic algorithm. Initially, the genetic algorithm with ABC operator will extract the best attributes and based on the extracted attributes the CHAID will generate the decision tree. We analyze the performance with different datasets and compare the analysis with the existing technique.