Ensemble Approach for Early Prediction of Malnutrition Level of Children: A Case Study on Children Under Five Years Old

Malnutrition refers to inadequate nutritional intake, nutrient imbalances or impaired nutrient utilization which is a global health issue that has a significant effect on children, leading to short-term and long-term consequences for their growth, development and overall well-being. The critical period for a child's growth and development is 1 to 60 months. The early identification of a child's malnutrition status is a strong foundation for lifelong health and well-being. Therefore, the objective here is to develop a model for early detection, and prediction of malnutrition in children. This study draws on a sample of 574 children under the age of five years in the Lunugala area, who were chosen randomly to be involved in the development of a predictive model for malnutrition using ensemble learning techniques. After pre-processing the collected sample, the dataset was partitioned into 70% of training set and 30% of testing set. Nine Machine Learning algorithms, namely Support Vector Machine (SVM), Decision Tree (DT), Guassian Naïve Bayes (GNB), Bernoulli Naïve Bayes (BNB), Complement Naïve Bayes (CNB), Logistic Regression (LR), Random Forest (RF), AdaBoost and Extreme Gradient Boosting (XGBoost) were initially used. And the evaluation metrics including accuracy, precision, recall and F-measure were utilized to access the performance of these individual algorithms. After comparing the nine base learners with each other based on accuracy, the three best performing Machine Learning algorithms were combined and integrated into an ensemble model using techniques such as Soft Voting Classifier, Hard Voting Classifier and Stacking Classifier. The evaluation of these ensemble models, also based on accuracy, precision, recall and F-measure revealed that the Stacking Classifier achieved the highest accuracy of 93%. Consequently, the study concludes that the ensemble learning technique is well-suited for better prediction of malnutrition status.