COVID-19 impact: Customised economic stimulus package recommender system using machine learning techniques.

Area of Science:

• Utilizes data analytics and machine learning for economic policy recommendations.
• Applies predictive modeling to understand household financial assistance preferences.

Background:

• The Malaysian government implemented the Prihatin Rakyat Economic Stimulus Package (ESP) to mitigate COVID-19's economic impact.
• The ESP includes diverse financial aid types, but household preferences vary significantly.
• Customizing ESPs is crucial for effectively supporting low-income households.

Purpose of the Study:

• To design a recommender system for ESPs using data analytics and machine learning.
• To predict individual household preferences for different types of financial assistance.
• To enable customized ESP delivery for enhanced economic burden management.

Main Methods:

• Employed a dataset from the Department of Statistics Malaysia on COVID-19's economic effects.
• Applied the Cross-Industry Standard Process for Data Mining (CRISP-DM).
• Developed and compared four machine learning models (Decision Tree, Gradient Boosted Tree, Random Forest, Naïve Bayes) to predict preferences for moratorium, utility discounts, and EPF/PRS cash withdrawals, selecting the best based on F-score.

Main Results:

• Gradient Boosted Tree demonstrated superior predictive performance across all subsidy types.
• Achieved high F-scores: 87.6% for moratorium, 84% for utility discounts, and 82.4% for EPF/PRS cash withdrawals.
• Identified that households preferring moratoriums generally did not favor other aids, except cash assistance.

Conclusions:

• Developed effective machine learning models for predicting household ESP preferences.
• These models offer a pathway to designing personalized economic stimulus packages.
• Customized ESPs can significantly improve the management of financial burdens for low-income households.