Application of Data Mining Technique to Develop Chronic Disease Distribution Map using Drug Distribution Data in Ethiopia
No Thumbnail Available
Date
2013-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Abeba University
Abstract
Background: Due to the great difference in population structure, geographic environment, food composition, ethnicity and lifestyle, it could be predicted that there may be significant differences of chronic disease forms and distribution in the various administrative areas.
The amount of data getting generated in any sector these days is enormous. There are many data mining tools and technique to uncover hidden knowledge in the data. At the same time Ethiopian PFSA has huge and useful drug distribution data in their data base to investigate chronic disease distribution.
Objective: The purpose of this study is to investigate distribution of chronic diseases in various administrative areas of the country based on chronic disease drug distribution data applying data mining techniques.
Methods: Drug distribution data was collected from EPFSA. Data that are retrieved from the organization is from 2003 up to 2005 EC. Since annual data follow is high and distribution density is the same, two and half year’s data is enough to produce distribution map and identify increase in demand applying data mining technology. Any data beyond these years are redundant and over saturate the models.
In order to optimize the desired outcome the researcher has followed Hybrid data mining process model.
The model is selected because it is appropriate for academic research; it combines the best features of KDD and CRISP; and starts with problem domain understanding.
Results: The study revealed that some drugs are more important at one hub than the other.
Gullele hub received the hieghtest persontage of Athma (17.3%), Cardiac (38.5%), Diatetes (45.6%) and Hypertenion (28.99%). While Parkinson drugs are issued mostly to Mekele (15.5%)hub.
The mining software revealed that some drugs are more important at one hub than the other in specified time.
Conclusions: Issue date, issue number and expiry dates are selected as best attribute by the mining tool. Based on discussion with domain experts issue date is important for drug distribution while issue number and expiry date are not relevant to the drug distribution.
Description
Keywords
Data mining,disease,map,drug