A Hybrid CBR Classification Model by integrating Decision Tree and Random Forest into Case Retrieval

Abstract

Due to the huge amount of medical data, which are stored in databases. The classification is the most demanding task for automatic decision making. This paper presents the development of ahybrid system for classifying medical data, based on the combination of learning methods with Case-Based Reasoning (CBR). The importance of this work lies in the design and implementation of an automatic classifier such as decision trees (C4.5,REPTree, LMT) and Random Forests (RF) to model the Retrieval phase of a CBR system, thus aiding inthe diagnosis or initial screening of the disease. The performance of the hybrid CBR system designed with C4.5 is compared to those designed with REPTree, Logistic Model Tree (LMT) and Random Forest (RF) for a set of dynamic and static activities. The system is trained and tested on four medicaldatasets, namely Wisconsin Breast Cancer, Thyroid, Hepatitis and Breast pathologies. Simulation results show that the proposed CBRRF and CBR-LMT methods outperform CBR-C4.5 and CBR-REPTree by achieving overall accuracy during cross-dataset evaluation of 97%, 96%, 83%, and 94%on Wisconsin Breast Cancer, Thyroid, Hepatitis, and Breast pathologies, respectively, and achievebetween0.10-0.14 root mean square error.