Time series analysis tools for forecasting technology stock prices

Abstract

Active investors and fund managers are constantly trying to forecast the price movements of assets making up their investment portfolio. Being able to deduce such information is key in spotting market trends, leading to taking optimal investment decisions and ultimately maximizing one’s wealth. This paper attempts to predict stock prices of highly-capitalised companies listed on the NASDAQ index, which operate in the ever-growing technology sector. The forecasting techniques used rely on historic prices and are classified as univariate time series forecasting models, including the Random Walk, Exponential Smoothing, ARIMA and ARIMA-GARCH models. When comparing out-of-sample forecast error metrics, such as the MAE and RMSE, we find that the non-linear ARIMA(1,1,0)-GARCH(1,1) is the best model for producing short-term out-of-sample forecasts in the technology industry.