Design and Implementation of Expert Clinical System for Diagnosing Diabetes Using Data Mining Techniques

Abstract

Objective: The aim of this paper is to design and implement an expert clinical system to diagnose the type of diabetes and the levels of risk among diabetic patients using the data mining techniques clustering and classification. Methods: The research design made use of primary and secondary data and the data were collected using data collection tools and techniques such as questionnaires, direct interview and survey of existing medical records from 650 diabetic patients. The study was based on purposive sampling type using a structured questionnaire that was pre tested with 25 respondents. After making necessary modifications from the feedback received from the pre-test, the final questionnaire was prepared. Findings: With six iterations the data could be successfully clustered into three clusters namely type-1, type-2 and gestational diabetes using Simple K-means algorithm. The classification algorithms - NaiveBayes, Random Tree, Simple Cart and Simple Logistic were used on the clustered data to classify the data into mild, moderate and severe types resulting into an expert clinical system. Conclusion: This paper demonstrates creation of expert clinical system for the diagnosis of the diabetic mellitus using clustering and classification techniques of data mining. However with suitable modification the same can be extended to evolve similar systems in other application areas in health care.