Predicting Malnutrition Status of Under-Five Children in Dhamar Governorate, Yemen Using Data Mining Techniques

Abstract

Malnutrition is characterised by the insufficient intake of certain nutrients and the inability of the body to absorb or use these nutrients. This health problem keep going to be a real challenge among children under five years of age in developing countries, including Yemen, despite good aids provided. So, malnutrition is a health problem that significantly participates to child mortality rate in Yemen. The overall prevalence of malnutrition among children in Dhamar Governorate has significantly higher rates compared to other Yemeni governorates.
 In this paper, an intelligent predictive system using data mining classification techniques such as J48 decision tree, Bagging and Multi-Layer Perceptron Neural Network (MLPNN) for predicting malnutrition status of under-five children in Dhamar Governorate is proposed.
 The main objective of the present paper is to study these classification techniques to predict the 2018-2019 Dhamar Governorate, Yemen Demographic and Health Survey (DGYDHS) dataset and find an efficient technique for prediction. This dataset is imbalanced, so Synthetic Minority Over-sampling TEchnique (SMOTE) is utilised to balance the dataset.
 The obtained results were evaluated by the famous performance metrics like Accuracy, TP (True Positive)-rate, FP (False Positive)-rate, Precision, F-Measure, Receiver Operating Characteristics (ROC) graph and execution time. The obtained results revealed that the three classifiers with all attributes have higher predictive accuracy and are generally comparable in predicting malnutrition cases.