An intelligent support system for diagnosing dehydration in children

Abstract

This paper present the implementation of artificial intelligent on medical decision support system for diagnosing childrens dehydration. In this study, the intelligent system constructed using decision tree method with C4.5 algorithm and pruned with REP (Reduced Error Pruning) method. This study was a collaboration between the doctor and the hospital in order to analyze the dataset of children dehydration in Indonesia. The number of 92 medical data was recorded for dataset and divided into two subsets: trainingset (57 records) and testset (35 records). The medical symptoms of dehydration that used for Input variables are general appearance, eyes, respirations, turgor and mucous membranes, while the output variable is the severity of dehydration that classified into three categories: severe dehydration, some dehydration and no dehydration. The validation was done by comparing the classification performance of the intelligent system and the doctor diagnose. The confusion matrix was used for mapping the classification performance of intelligent system and evaluated by using accuracy and the value of error rate. The result show that, the implementation of artificial intelligent on medical decision support system have an accuracy of 91% and the error rate value of 0.085714286. From the result it can be concluded that the implementation of artificial intelligent on medical decision support system can be use for supporting dehydration diagnostics in children.