Global Hunger Index: A multistage coefficient estimation analysis using machine learning techniques for a hunger free society

The Global Hunger Index (GHI) is one of the most prominent indices for calculating and describing the severity of hunger level in various countries. Nutrition, mortality, and food supply are the basic parameters defining the hunger index value. Using a single method for coefficient estimation of hunger index parameters is not a viable option. It is necessary to design a new framework which can help us to explore more possibilities for coefficient estimation and for identifying the critical features as well. In this paper, we have proposed a new coefficient estimation technique called Multi-Stage Coefficient Estimation (MSCE) strategy that enhances the prediction rate. MSCE is an aggregation of multiple methods for generating the coefficient values of the associated features group. Decision Tree Regression, Random Forest regression, Logistic Regression (comprising of lbfgs, saga and sag problem solvers), Simple Regression and Multiple Regression coefficient computing methods are used in the proposed framework to estimate the coefficient values. The obtained coefficients were then used to build the max, avg and min regression equations for prediction. The introduced methodology is applied on the global hunger index data collected from the GHI repository. The results of the proposed MSCE methodology were compared with other popular approaches like Simple Linear Regression, and Multiple Linear Regression in terms of Mean Absolute Error, Mean Square Error, Root Mean Square Error, R- Squared and Adjusted R-Squared metrics. The proposed MSCE strategy achieved better results when compared with existing methods.