The Author(s). Published by Elsevier Ltd on behalf of Surgical Associates Ltd. This is an open access article under th

Abstract

The nutrition may be well-nutrition or malnutrition. Malnutrition is a public health issue that affects a wide range of people of all ages, including a child, young, or adults, but it is more prevalent in preschool children. It is caused by both socio-economic and demographic factors. Knowing this factor and nutrition status was important in explaining the variation in the long-term nutritional status of children but the understanding of its factor and kowing of child nutrition status remains limited for both the patient and the doctor. As researcher’s knowledge, related works show that there is some limitation/gaps that attempt to predict nutrition status of children based on socio-economic and demographic factors using ensemble machine learning methods. In this study, hence, we develop the predictive model and prototype for child nutrition status. Data was collected from the Ethiopian demographic health survey from 2005 to 2019 G.C.. The data was pre-processed to get quality data that are suitable for the machine learning algorithm in order to develop a model that predicts the nutrition status of children under five years of age. The study was conducted by following a design science research approach. For constructing the proposed model, four experiments were conducted with a total of 44,484 instances with 25 attributes/features by splitting the dataset into training and testing dataset with split ratio of 80/20 using random forest, XGBboosting, adaBoosting, and CatBoost algorithms. The overall accuracy of random forest, XGBboosting, adaBoosting, and CatBoost with 25 features, with the accuracy of 86.19%, 87.47%, 90.28%, and 87.38%, respectively. The adaBoosting algorithm is selected as the best algorithm due to the result of both objective and subjective evaluation-based metrics. Hence, the model created by the adaBoost algorithm was used to identify determinant factors based on feature importance. Accordingly, the most determinant factors of child nutrition status are women’s age, source of drinking water, number of children under five, Region, place of residence, number of visits to the hospital, partner occupation level, religion, succeeding birth interval, size of a child during birth, duration of breastfeeding status, numbers of household members, mother educational level, and wealth index. These factors are considered to design the artifact using a flask framework. Once the artifact is ready, to simplify the demonstration and evaluation of the artifact Heroku-based cloud computing platforms are employed. The subject evaluation shows that the proposed artifact is able to register 80% of users’ acceptance. The developed model does not integrate with the knowledge-based system, therefore we recommend to future researchers integrate the model with the knowledge-based system so as to design an intelligent system that automates child nutrition status prediction. Keywords: - nutrition status, Children under