Authors : Eman Al-Shamery and Ali Rahoomi Al-Obaidi
Abstract: In the health care sector predict the type or severity of the diseases is important for helping people to know their health stability and find solution for any negative indicator. This study aims to improve prediction of the diseases by exploiting Modified Rough Set (MRS) for features selection which it is proposed as a new method and employing a Most Common Decision Tree (MCDT) which is suggested as a modified of decision tree method for making decision. In addition to pre-processing stage which contains the Mode-Relation-Average (MRA) method for filling missing value and grouping. The system consists of 2 main stages, first is features selection based on MRS and the second disease prediction using MCDT. The output of MRS Model is three subsets of features graduate according to the importance degree of features: Most Important (MI), Important (I) and Unimportant (UM). The MCDT is used to predict disease type. Subsequently, 3-cross validation is used in the testing process. The proposed system has been applied on binary and multi classes heart disease (5 classes). Finally, the proposed methods have been compared with the other methods such as Naive Bayes and Bayes Net. The Max correctly classified instances obtained using the proposed methods are better than 2 previous methods.
Eman Al-Shamery and Ali Rahoomi Al-Obaidi, 2018. Disease Prediction Improvement Based on Modified Rough Set and Most Common Decision Tree. Journal of Engineering and Applied Sciences, 13: 4609-4615.