Correlated gain ratio based fuzzy weighted association rule mining classifier for diagnosis health care data

Abstract

Healthcare data need an accurate diagnosis of diseases with the low computation time. Fuzzy association rule mining converts quantitative attributes to fuzzy attributes which maintain the integrity of information. Fuzzy association rule mining is most effective among various classification methods used for diagnosing health care data. The major challenge in the fuzzy association rule mining is to reduce the exponential growth of rules produced by fuzzy partitioning of attributes. The proposed method uses the correlation and gain ratio based average ranking feature selection followed by fuzzy weighted association rule mining classifier to diagnose the medical data set. The average ranking feature selection method improves classification accuracy and reduces the number of rules by ranking the proper potential attribute. The computation time is also minimized by reducing number of rules. The performance of fuzzy weighted association rule mining classifiers based on correlation and gain ratio average ranking feature selection is evaluated by comparing the classification accuracy with three classifiers and five benchmark data set collected from UCI repository.