IMPLEMENTATION OF CLASSIFICATION PREDICTIVE ASSOCIATION RULE ( CPAR) ALGORITHM TO DIABETES DIAGNOSE

Abstract

Abstract. Hospital database yielded from hospital information system generally contains very much data with various attributes. Filtering and presenting relevant infarmation in excessively database is difficult work. So needs certain techniques that screening of information can be done in efflcient and effective, for example by applying data mining which will trace patterns from data jar purpose of analysis. In this research studied how data mining can be applied to assist diabetes diagnose from data ofmedicat laboratory. The real medical data set concerns patients with diabetes mellitus risk are included in diabetes data warehouse. Three steps are implemented for data mining process building. The first step is to deal with missing values. Next is the discretization step, where each variable is divided into a limited number ofvalue groups. The next step is creating rule mining and classijication. There are 6.000 non-diabetes patients and 4.000 diabetes ones each with 12 variables: age, sex and results laboratory test. With Classification Predictive Association Rule ( CPAR) algorithm, maximum predictive accuracy for diabetes is 69% and non-diabetes is 81%' The decision ni/es furthermore can used in application to predict diabetes or not. Data mining system building using CPAR algorithm is useful to diabetes diagnose.