Factors Affecting the Use of Domestic Gas in Benin: A Comparative Study of Artificial Neural Networks and Logistic Regression

Abstract

The strong growth in demand for wood energy in Benin's major cities today represents a real threat to the preservation of forest ecosystems. The promotion of new alternatives such as the use of domestic gas as cooking energy could help to better cope with the adverse effects of climate change resulting from deforestation. The objective of this paper is to analyze the determinants of domestic gas use in Benin. To do so, we used data from 15,000 households collected during the Global Food Vulnerability and Security Analysis Survey of 2017. We then compared the prediction of household gas use determinants by Multilayer Perceptron Neural Networks (MLP) and classical Binary Logistic Regression (BLR). The two approaches have highlighted as important factors of the adoption of Domestic Gas in Benin, the residence department (here department of the Littoral) and the level of education. We also noted that the MLP highlighted more adoption factors than the BLR model (income, ethnicity, and number of wives of the household head). In order to increase the use of domestic gas on a large scale, the Government must put in place a policy that promotes the physical and financial accessibility (through subsidies) of the product to the large mass of the population in our cities which are still dependent on traditional energy sources such as wood fuel and charcoal in order to better protect our forest ecosystems in a sustainable manner. The Government could also strengthen the public-private partnership in this sub-sector by, for example, creating facilities for private economic operators through tax or customs exemption measures.