Data Mining Techniques to Diagnose Diabetes Using Blood Lipids
|
Reza Rafeh * 1, Mohammad Arbabi2 ![](./files/0allsites/images/pubmed20.png) |
1- Arak University , r-rafeh@araku.ac.ir 2- Islamic Azad University, Malayer Branch |
|
Abstract: (9256 Views) |
Introduction: Nowadays, diabetic disease is one of the most common, dangerous and costly diseases in the world spreading rapidly. Data mining techniques can be used for early diagnosis of this disease which results in preventing a lot of problems for patients including heart diseases, vision problems and kidney disorders.
Matherials & methods: In this research, the Rapid Miner software has been used as a modeling tool to classify each patient as either diabetic or non-diabetic. The data set of this research has been collected from the database of one lab in Nahavand which includes the information of 5706 patients in a five years period from 2009 to 2013. The data set includes such information about patients as: age, gender, the level of lipid in the blood and the amount of fasting blood sugar.
Findings: After modeling with different classification techniques, the best accuracy achieved from the decision tree c4.5 which was 90.02%.
Discussion & Conclusion: For early diagnosis of diabetes in many countries around the world many techniques have been proposed using a variety of methods and variables. In the current research, using the relationship between blood lipids and fasting blood sugar, a method based on data mining techniques for diagnosing diabetes has been proposed. |
|
Keywords: Data Mining, Diabetes, Classification techniques, Decision tree C4.5. |
|
Full-Text [PDF 476 kb]
(11067 Downloads)
|
Type of Study: Research |
Subject:
stats Received: 2015/01/2 | Accepted: 2015/04/12 | Published: 2015/12/19
|
|
|
|