Developing A Predictive Model For Pre- Diabetes Screening By Using Data Mining Technology
No Thumbnail Available
Date
2017-06
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa University
Abstract
Introduction - Diabetes is one of the most common non-communicable diseases. That has a
significant contribution of increased morbidity, mortality and admission rate of patients in both
developed and developing country. The burden is also very enormous in Ethiopia with estimated
1.4 million in World Health Organization country profile report (2014); even this doesn’t
included pre-diabetes and undiagnosed cases. International Diabetes Federation report an
estimated 83.8% of all cases of undiagnosed diabetes mellitus are in low- and middle-income
countries. Therefore early screening, diagnosis and prompt treatment are needed to prevent
comorbidity and mortality, delay the onset of disease, and reduce serious complication and
permanent damage.
Objective: The aim of this study was to develop a predictive model for screening of pre-diabetes
patient using data mining technology.
Method: This study conducted in Adare general hospital in Hawassa city, south Ethiopia. The
methods used for mining, Cross-Industry Standard Process of Data Mining which contains six
phases such as problem understanding, data understand, data preparation, model building,
evaluation and deployment was used. In general, 4529 of age > 20 years visiting diabetic unit
for general medical examination and follow up were included from January to March 2017.
Designed template was used for data collection. For data pre- processing was used Microsoft
Excel and WEKA open source software for mining.
Results and discussion: - The study has revealed that the model constructed PART with all
attributes registers the highest accuracy of 96.78% as compared to J48 decision tree which was
93.66%. The finding of the study clearly presents that screening of diabetes and pre diabetes
patient. Based on result of prediction designed project prototype model that predict whether the
positive risk of diabetes or not based on this result patients should link further investigation or
provide council for future the way to prevent or delay on set of diabetes.
Conclusion: - Generally, the prototype system serves as a guideline, diabetic screening to
support early detection of patient. The initial feedback from health works has been extremely
positive. Hence the developed prototype system achieves a good performance and meets
the objectives of the project.
Recommendation: - based on finding of project forwarded recommendation for respective
stockholders.Develop diabetes screening model by using data mining technology
Description
Keywords
Develop diabetes screening, model by using data mining technology