A research submitted in partial fulfi llment of the requirements for the degree of Master of Science in Information Management, School of Engineering and Technology

Data mining techniques are resu lts of a long processing of anal ysis, prediction, research and platform based development. Modern medical institutes and techniques deal with huge amount of information stored in medical databases. The effective practice in medical mining is achieved through improved medical diagnosis using automatic analysis techniques from patient{u2019}s details and case history stored in medical records. Diabetes (Diabetes mellitus) is a chronic diseas e with 382 million people affected with this disease. There are three type of Diabetes namely T ype 1, Type 2 and Gestational. The datasets for this study consisted of 4,104 diabetes patients (bot h IPD and OPD) from the Phra Pok Klao Chanthaburi Hospital of Thailand between the periods of 2009 to 2013. The main purpose of this study is to analyze risk of diabetes in patients and predict death risk using classification techniques of data mi ning depending upon many factors and obtained medical datasets. This study also aims at discovering role of da ta mining tools and techniques for massive medical data evaluation that provides po werful assessment for medical decision making. The comparisons of the different classifiers and visualization through different decision tree techniques have been applied. The results of the stu dy determined some crucial finding s for the purpose of prognosis of diabetes disease. Type 2 diab etes was common among the patients with various complications (respiratory organs, abdominal or gans, heart, blood cells, skin a nd urinal organs) leading to death, DM risk and long span duration of the dis ease. Diabetes Mellitus (DM) risk tended to be higher in patients with age of 31 to 60 and 61 to 100 on basi s of HbA1c test results and classification. Some factors like age, HbA1c test, comp lications, type of diabetes, drug type illustrated high significance effect in predicting DM risk and death risk due to diabetes suffering. However, no significance of DM risk based on gender was found.Therefore, the significance of the medical analys is depended on the dataset obtained that consisted of diabetic patient{u2019}s diagnosis hist ory and the chosen classification model namely Naïve Bayes classifier along w ith decision trees generated by CHAID and Random tree.

Research Studies Project Report (M.Sc.) - Asian Institute of Technology, 2014