The foreign exchange (FX) market is long established as the largest and most importantglobal financial market. While a large number of research papers focus on forecastingin the FX market, there are still gaps in the literature. First, very few papers focus onimproving the parameter estimation process in the forecasting context. Second, artificialneural networks (ANN) with large sizes have not been applied to FX forecasting with therecently-fast-developed GPU techniques. Third, forecasting for trading purposes in theFX market has been limited to either building forecasting models or analysing technicalindicators. A combination of ANN forecasting models with technical indicators is rarein the existing literature. The use of more-accurate parameter estimation algorithms andGPU techniques also makes the thesis unique in the methodological sense.The thesis uses three types of ANN models, namely GARCH-ANN, large MultilayerPerceptron (MLPNN) and Long Short Term Memory (LSTM), to forecast volatility, thedirection of price movements and price patterns in the FX market. Research is conductedat three data frequencies, namely monthly, daily and hourly as the analysis goes from themacro-perspective to the micro-perspective'
In the first empirical chapter, a Recursive Simulation Algorithm (RSGA) is proposed forestimating the parameters of a volatility forecasting model using GARCH-ANN. Theproposed algorithm significantly improves the stability and accuracy of the estimationprocess by dealing with the local-optimum and convergence problems. The secondempirical chapter utilises a large MLPNN model with GPU implementation to forecastthe price direction of different FX pairs, with over 40 macro-economic indicators as inputvariables. Highly profitable out-of-sample results are observed for some of the currencypairs, which challenges the semi-strong form of the Efficient Market Hypothesis (EMH).Significant efficiency improvement is achieved with the GPU implementation. The thirdempirical chapter proposes the use of the Relative Strength Indicator (RSI) as a measure of the extent of trend-following and mean-reversion patterns of FX rates. A LSTM modelis utilised to forecast price movement patterns (measured by RSI). The trading strategybased on forecasting results of price movement patterns generates more stable profitsthan the benchmark Moving Average (MA) or RSI implemented on their own. However,the overall low profitability over time for the four currency pairs fails to challenge theweak-form EMH.Overall, with the novel methodologies and technologies implemented within differentmodels, this thesis finds evidence on some extent of inefficiency of the FX market atlower trading frequency (e.g. monthly) and less inefficiency of the FX market at highertrading frequency (e.g. hourly). One possible explanation is that at higher frequencies,the large number of daily (or higher frequency) traders and high-frequency tradingalgorithms reduce both the number of mis-pricing opportunities and the length of timethat any mis-pricing opportunity may last.