Modeling And Predicting GDP Growth Using Univariate Time Series and Machine Learning Methods: A Comparative Analysis of Pakistan, India and Bangladesh.

Abstract

Accurate forecasting of GDP growth is vital for economic planning and policy-making, especially in developing countries such as Pakistan, India, and Bangladesh. This study conducts a comparative analysis of nine univariate time series forecasting models to predict GDP growth rates for these three countries using annual data from 1961 to 2023. Forecast accuracy is evaluated using six metrics: RMSE, MAE, MAPE, NSE, KGE, and R². The results demonstrate that the Neural Network Autoregression (NNAR) model consistently outperforms traditional and other statistical models, effectively capturing nonlinear trends and fluctuations in GDP growth. These findings underscore the limitations of conventional models for volatile economic data and highlight the potential of advanced and hybrid modeling approaches, including machine learning and deep learning techniques, to provide more accurate and reliable GDP forecasts.

Keywords:

GDP growth, time series forecasting, ARIMA, NNAR, machine learning, economic prediction