A Data Mining Framework for building Dengue infection disease Model

Abstract

Dengue infection is an epidemic disease typically found in tropical region. Symptoms of this infection show rapid and violent to patients in a short time. There are 4 classes of Dengue infections which are DF, DHF I, DHF II, and DHF III. Nowadays, the experts need to know the set of features on dengue infection in order to correctly classify the patients. Our temporal dataset consists of clinical data and laboratory data. The data was collected from the first visit of patient until the date of discharge. We obtained 3 datasets from different regions of Thailand which are Srinagarindra Hospital (KK: 440 patients), Songklanagarind Hospital ( SK : 330 patients) and Siriraj Hospital (SR: 258 patients). Each dataset consists of more than 400 attributes. The second objective of this research is to detect the day of defervescence of fever which is called day0. The day0 date is the critical date of Dengue patients that some patients face the fatal condition. Therefore the physicians need to know the feature sets, those have effect on the condition. They expect to have an intelligent system that can trigger the day0 date of each patient. To accomplish the knowledge discovery task, we consider to employ decision tree as a data mining tool. We propose a set of meaningful attributes from the temporal data. We analyzed the result of dengue's decision tree and day0's decision tree in discussion part. Finally, we obtained high accuracy (97.0 %) and we got the new set of features that can be applied to real world data.