Using random forest (RF) to identify key determinants of poverty in Kenya.

Abstract

Background The World Bank (WB) has defined poor people as those who live on USD 1.9 or less per day. Most of the poor populations live in sub-Saharan Africa, Kenya being one of them. Although Kenya has made tremendous improvements in poverty reduction, it is unlikely to eradicate it by the year 2030. Its on this backdrop that it is important to understand the determinants of poverty in Kenya. This study intends to apply a random forest technique to KDHS 2014 dataset to explore poverty determinants in Kenya. Methods Random Forest is an algorithm used for classification and regression usually constructed from a set of classification and regression trees. The advantage of random forest is that they perform well in large feature setups and is effective in handling complex data. Analysis is carried out using the random forest package in R software, while imputation is carried out in miss Forest package. Inference is made on the model’s classification accuracy, model diagnostics and interpretation. Results The results showed that variables had a very small percentage of missingness. There was reduction in error of the classifier as the trees in the classification was increased. The variable with the highest importance was the highest education level attained while marital status and sex of the household head, were variables with least importance. The classification was improved on the extreme ends of the wealth index as the Random Forests highly classified poorest and richest classes, while the middle, poorer and richer indices were not as accurately classified. Conclusion This study brings out the application of random forests to classify and predict the wealth index class of an individual in Kenya. The random forests are a significant improvement from classical regression techniques. The multiclass classification problem was effectively captured in this study. Regional residence and level of education details should be considered when interventions are being made for improvements of livelihoods in the country.