Optimizing Multi-Path Decision Tree by Clustering and K-Nearest Neighbor Methods

Binary decision trees are not adequate to make the decisions more factual and lucid; rather we have to renovate this technique so that it can congregate the real world postulation and be able to support the decision making even in case of extraneous and mislaid data. This paper develops a classifier which is actually a pruned multi-path decision tree based on agglomerative clustering and K-Nearest Neighbor (K-NN) methods. Further, it states the advantages of this classifier over the binary decision trees and certain assumptions for the proposed method. The paper also demonstrates how the missing values and outliers can be found out to optimize the method