Study of classification rules on weighted coronary heart disease data

This paper studies the data weighting affection of rules that derived from different decision tree algorithms. The data that we study is collected from patients with coronary heart disease; it has 1723 records and 71 attributes in each record. The weighting methods we used are system reconstruction method and principle component analysis, from which we get two weighted data sets. The analysis based on the weighted data sets and the original not weighted data set. The analysis methods we adapted are some decision tree algorithms; they are C4.5, CART, CHAID and Exhausted CHAID. In the analysis, we compare the quantity and quality of rules which are proved correct by medical specialists and clinical doctors. According to the result from experiments, we know that weighted data sets can improve rule quality, which is better than the result from not weighted data set.