Comparison Of Machine Learning And Time Series Models For Economic Growth Forecasting: Empirical Evidence From Pakistan
Author: Ali Asgher

The aim of this study is to compare machine learning and time series models for economic growth forecasting. The economic growth forecasting was analyzed based on export of goods and services, import of goods and services, trade openness, exchange rate, inflation, unemployment, remittances inflows, gross fixed capital formation and foreign direct investment. Machine learning (SVR and ANN) and time series (ARDL, AR, RW) model were used to forecast economic growth. To compare the forecasting performance of machine learning and time series models, root mean square error (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE) were used. Based on empirical evidence from Pakistan, using quarterly time series data from 1981 to 2019, the key findings of this study are that all the models perform well but ARDL model forecast economic growth more accurately than all other machine learning and time series models. Given the results, ARDL model can be applied effectively in the applications of economic growth forecasting. Supervisor:-Dr. Hafsa Hina

Meta Data

Supervisor: Hafsa Hina

Related Thesis​