Stock market predictions based on quantified intermarket influences

This study aims to make predictions about the Australian All Ordinary Index (AORD). The following two types of predictions are considered: (1) predicting the direction (up or down) of the Close price; and, (2) predicting whether it is best to buy, hold or sell. A novel approach, which heavily involves global optimization, is adopted for predictions. This thesis investigates different methods of incorporating intermarket influence for predictions. It proposes a novel method for quantifying stock market influence from a set of potential influential markets on a given dependent market, by maximising the rank correlation between the markets of interest. The possible intermarket influence from the world?s major stock markets on the AORD was quantified using this method. The ways of using quantified intermarket influence for the predictions were investigated. The direction of the Close price of the AORD was predicted using the Feedforward neural networks (FNNs). When predicting whether it is best to buy, hold or sell, to overcome the difficulties caused due to the imbalanced distribution of data (as a result of considering the hold class), this thesis introduces some neural network algorithms. These new algorithms use modified error functions and were trained with a global optimization algorithm. The results relevant to both types of predictions suggest that the quantified intermarket influence on the AORD can be effectively used. This is an indication of the effectiveness of the proposed approach for prediction.