Abstract:
In this paper, we considered migration induced predictions by machine
learning algorithms. In the determinants of migration, which are age,
occupation, psychological pressure and social media effect, the costs are
known. The models are trained on a full data set prior to launching. It applies
Random Forest, K-Nearest Neighbours (KNN), Support Vector Machines
(SVM), XGBoost and Logistic Regression methods to migration tendency
prediction. The Random Forest model (best performing) also made high
accurate predictions, 81.35% correct predictions also. It also contains a
recommendation module for personalized feedback to subjects. Last but not
the least, `model interpretability methods’ like LIME and SHAP are used to
serve the prediction in an interpretable way. So from that standpoint it’s an
experiment for the first time and it may work out.
The research ethical considerations involved, such as data privacy, fairness
and interpretability, could be said for the opposite case which occurs as the
integrity for machine learning can be strengthened. Offering us predictive
analytics and implementable solutions to these challenges, it is improving our
migration process and emerging as more one governed by public decisions. It
offers scalable, interpretable predictions for migration and new directions of
research in this domain.
