The stability of prices is an important indicator of overall economic performance and is one of the main objectives of monetary policy. Pakistan's economy has a long history of unstable macroeconomic performance, especially the persistence of high inflation rates, which lasted for almost three decades. During this long period, many stability programs mostly backed by the IMF could not be implemented thoroughly and failed to achieve the desired outcomes and price stability. Researchers have identified factors including firm size, past stock performance, value, and growth as some of the factors affecting the stock exchange. While the current study uses only the Karachi Stock Exchange 100 (KSE100) index as a proxy and performs a time-series analysis to identify the best forecasting model which can help investors and government agencies to make up-to-date decisions. Recently, forecasting future observations based on time series data has received great attention in many fields of research. Several techniques have been developed to address this issue to predict the future behaviour of a particular phenomenon. In this study, two methodologies for forecasting the KSE100 index are used. The first is the linear time series modelling consisting of NAÏVE and Box-Jenkins methodology, while the second is the non-linear methods consisting of Generalized Autoregressive Conditional Heteroskedasticity (GARCH), Artificial Neural Network (ANN), and Support Vector Machine (SVM). These two approaches are used to obtain the static and dynamic forecasts of the KSE100 index daily data, and the accuracies are compared by using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Directional Statistics (DS) and Diebold-Marino (DM) Test. The results indicated that the ANN is the most effective machine learning approach for improving the forecasting accuracy of the KSE100 index. Thus, from this study, the recommended model for forecasting the KSE100 index data is ANN which can handle the non-linearity and non-stationarity. Hence, the ANN model is recommended and could be used for forecasting the KSE100 index data.
Key words: Econometrics, Forecasting, KSE100 Index, Machine Learning, Time Series
|
