GDP prediction of The Gambia using generative adversarial networks

Area of Science:

• Economics
• Data Science
• Machine Learning

Background:

• Accurate Gross Domestic Product (GDP) prediction is vital for national economic analysis and growth.
• Traditional forecasting methods may struggle with complex economic interdependencies and limited data scenarios.
• The Gambia's economy presents a case study for advanced predictive modeling due to its specific characteristics.

Purpose of the Study:

• To forecast GDP using key economic indicators including government spending, inflation, aid, remittances, and FDI.
• To evaluate Generative Adversarial Networks (GAN) as a deep learning approach for enhancing GDP prediction accuracy.
• To demonstrate GAN's effectiveness in small data environments for economic forecasting.

Main Methods:

• Implementation of Generative Adversarial Networks (GAN) for GDP prediction.
• Utilizing economic factors such as government spending, inflation, official development aid, remittance inflows, and Foreign Direct Investment (FDI).
• Comparative analysis against Random Forest Regression (RF), XGBoost (XGB), and Support Vector Regression (SVR) models.

Main Results:

• Generative Adversarial Networks (GAN) achieved the highest prediction accuracy at 99%.
• RF and XGBoost models also demonstrated strong performance with 98% accuracy.
• GAN proved superior in capturing intricate correlations between GDP and its influencing economic factors.

Conclusions:

• Generative Adversarial Networks (GAN) offer a highly accurate and desirable method for GDP forecasting, especially in economies with limited data.
• The study provides a valuable tool for policymakers and stakeholders to develop strategies for sustained economic growth.
• Accurate GDP predictions derived from advanced machine learning techniques can support informed economic decision-making.