Reverse migration prediction model based on machine learning / Azreen Anuar

Reverse migration in Malaysia is a relatively new emerging phenomenon where migrants intentionally choose to return to their hometown for better living. Thus, there is a demand to investigate the determinants that lead to these changing population mobility trends in Malaysia. Migration predictions are notorious for bearing high error rate because migrations are the most complicated and unpredictable of the key demographic processes. A significant way to minimize the errors is by using a machine learning approach that can predict reverse migration intelligently depending on the tested dataset. Thus, this research aim to develop a reverse migration prediction model based on machine learning. To fulfil this aim, this research proposed three (3) objectives. The first objective is to identify the factors influencing reverse migration based on the statistics from previous empirical studies through a systematic literature review. The second objective to analyse the relationship among the factors that influence reverse migration in Malaysia using empirical experiments performed through the Shapiro-Wilk and Spearman Correlation analysis. And the third objective is to evaluate reverse migration prediction model based on machine learning analysis. For this purpose, three (3) algorithms have been assessed, namely, the Random Forest, Decision Tree, and Gradient Boosted Tree. The findings of this research have provided new insights into the six (6) factors that could influence reverse migration. In addition, the results from the three (3) algorithms that were tested showed that Random Forest outperforms other algorithms by acquiring an accuracy and classification error to predict reverse migration. With the application of machine learning aligned with Industrial 4.0, this research would be advantageous to predict reverse migration in a more efficient way.