Hybrid mechanism for medical data classification

Abstract

In recent years, medical data mining is the hot topic of research. Earlier extraction was done manually so the process was time consuming. Thus, to solve this drawback machine learning technique was implemented to extract knowledge from the large dataset. In this paper, the medical data processed for removal of missing values, then orthogonal local preserving projection (OLPP) algorithm is employed for feature dimension reduction. Then, fuzzy rules are generated and optimised by binary cuckoo search algorithm. Further, the classification is carried out using the generated fuzzy rules and the best rule is selected based on the accuracy using the decision tree classifier. The performance of the proposed technique has been evaluated using University of California Irvine (UCI) datasets namely Cleveland, Hungarian, Mammographic masses, PID and Switzerland. After observation and comparison, the method achieves highest accuracy of 97.3% for PID dataset whereas the existing method provides 89.5% accuracy for PID dataset.