Tree-based machine learning in classifying reverse migration/ Azreen Anuar, Nur Huzeima Mohd Hussain and Hugh Byrd

Reverse migration is an increasingly urgent issue as it is influenced by various factors such as economic crises, political turmoil, natural disasters, and the COVID-19 pandemic. Predicting reverse migration can provide valuable insights for policymakers and stakeholders to design appropriate interventions. However, there is a scarcity of studies that have applied machine learning algorithms to this problem. This paper aims to fill the gap in the literature by discussing the application of machine learning algorithms for predicting reverse migration. The study compares the performance of three types of treebased machine learning (Decision Tree, Random Forest, Gradient Boosted Trees) with linear-based algorithms (Logistic Regression, Fast Last Margin, Generalized Linear Model). In addition to accuracy, this study also measured the area under the curve (AUC) metric, which has been seldom explored in previous research of reverse migration prediction. The findings revealed that tree-based machine learning algorithms performed slightly better than linear-based algorithms in terms of accuracy of prediction, with an improvement of approximately 1%. Based on the accuracy and AUC results, Gradient Boosted Trees is selected as the best algorithm. The findings of this study suggest that machine learning can provide valuable insights into predicting reverse migration. With the use of appropriate machine learning
algorithms, policymakers and stakeholders can make more informed decisions to address the challenges posed by reverse migration.