Description
A project presented to the College of Engineering in partial fulfillment of the requirements for the degree of Master of Science thesis in Engineering Systems Management by Mohamed Ismail Shaban entitled, "Risk in Forecasting Correct Directional Changes in Financial Time Series," submitted in May 2012. Master's project advisor is Dr. Hazim El-Baz. Available are both soft and hard copies of the project.
Abstract
Forecasting future moves of financial time series is an important task. Decision makers use different forecasting approaches. Regardless of the approach used, decision makers need to confident that the forecasting approach and methodology they use is the most appropriate one for the financial time series they are analyzing and the risk associated with it. A financial risk in this research is defined as the probability that the forecasting method will not predict the Correct Directional Change of the next periods' move. The approach of this research aims at analyzing the error in predicting the correct directional move from period to period. The accuracy will be determined by the ability of the forecasting method to predict the correct move (up or down) of the next day, week, and month. Six forecasting methods are evaluated in this research: Naïve forecast, 14 Periods Moving Average, ARIMA, Exponential Smoothing, ARRSES and Neural Network. These forecasting methods were used to predict the closing value of the Next-Period's move as well as its direction. A new approach is also presented in this research based on which decision makers may use to compare the forecasting accuracy and associated risk of different forecasting methods to predict the Correct Directional Change of the next periods. The use of the new approach is illustrated by applying it to four financial time series, which represent different financial markets namely: Spot Oil Price, Gold Daily Price, FTSE100, and Euro to Dollar Daily Exchange Rate. The results of the four markets indicated that using any of the six forecasting methods will have a success percentage of approximately 50% with small variation form a forecasting method to another. In addition, when comparing the six forecasting methods using the new presented approach, the 14 Periods Moving Average has proved to be the least risky forecasting method in three out of the four markets which are: Brent Spot Oil Price, Gold Daily Data, and Euro to Dollar Exchange Rate while Neural Network has proven to be the least risky in forecasting the FTSE100 stock market. Search Terms: Risk, Forecasting Methods, Correct Directional Change