Early Childhood Deaths Prediction using Machine Learning and VAR Model

Abstract

Health-related issues are always very serious. From 1st years there are always policies have been set to improve the health and education of individuals. This issue is more serious in developing countries. One of the problems of infant mortality due to lack of facilities, medicine, and financial concerns are key challenges in various parts of the world. All of these are counted under the area of standard development goals (SDG). International organizations such as the World Health Organization (WHO) are always making efforts to improve life standards every year. They always generate health reports publicly so that the annual improvements can be analyzed and new reforms can be made. The serious issue identified for children is the infant mortality rate, as well as child casualties, particularly under 5 years old. There could be many reasons for child mortality or infant death. They include socioeconomic factors. This issue is serious in both developing and non-developing countries. However, countries facing poor economic conditions are suffering more. The proposed study includes artificial intelligence(AI)-based algorithms under supervised learning, specifically using Naïve Bayes (NB) and XGBoost algorithms, which are considered highly efficient for this domain to identify the child death rate under age 5. Another statistical model was used for the comparative study. The third algorithm is the Vector Auto Regression (VAR) model, which is also famous for identifying regressive patterns in the data. The dataset was split into training and testing subsets, employing data balancing techniques such as Synthetic Minority Over-Sampling technique (SMOTE) for qualitative data generation. Machine learning classifiers, naïve Bayes, and extreme gradient boosting were used for the results deduction. A comparative analysis via a confusion matrix was performed for the performance evaluation. The results reveal the chances of health impairments when a child belongs to a certain statistical frame. The accuracy and precision of the results indicate the performance.