Random Forest Prediction Model for Household Cooking Energy Sources in Nigeria

Abstract

The use of dirty fuel for cooking is one of the causes of CO₂ emission which adversely affects climate change. Due to availability and cheapness, most of the Nigerian households are using these dirty energy sources for cooking and other households’ use. The aim of the research is to develop a Machine Learning (ML) model to determine households’ cooking energy sources. In this study, a Random Forest model was developed for the classification of households’ cooking fuel sources in Nigeria. The study dataset was sourced from the National Bureau of Statistics (NBS) database for Nigerian General Household Panel Survey (NGHPS). A dataset from 4,980 households was preprocessed and most relevant features were selected using the Univariate feature selection technique. The Random Forest model was developed with a varying number of ensemble trees. Classifier Model trained using 200 ensemble trees outperformed the other models with 97% accuracy, 96% precision, 95% and 96% for recall and F1 Score respectively. This performance shows a significant improvement when compared to the model developed using the traditional decision tree model in previous study having an accuracy of 96%, precision of 95%, recall and F1 Score of 93% and 94% respectively. Having seen that the use of ensemble learning is capable of improving the model performance, in future, this study can be improved by hybridization of multiple promising Machine Learning algorithms.