Deep Learning-based Time Series Models for GDP and ICT Growth Prediction in India

Abstract

Approximately 13% of the country’s gross domestic product (GDP) depends on information and communication technology (ICT), and India’s digital economy accounts for nearly ${\$}$200 billion in economic value annually. The literature has well established the function of ICT in stimulating economic growth. The work focuses on the relationship between the ICT use index and GDP based on the last 30-year time series multivariate data. The variables having a high correlation with respect to GDP has taken for the analysis. This work also forecasts the next 10-year growth in ICT use index and GDP w.r.t each other using machine learning and deep learning models which are linear regression(LR), random forest(RF), temporal convolutional network(TCN), Kalman forecaster(KF), Neural Basis Expansion Analysis for Interpretable Time Series Forecasting (NBEATS) model and transformer model. There are eight performance metrics we have used for model evaluation. The transformer model has been suggested as the best predicting model with the least root mean squared error value of 0.326, mean absolute error of 0.51, mean absolute ranged relative error of 52.592, and so on. This paper also suggested policies to foster the country’s economic growth using ICT.