Using neural network prediction to arbitrage the Australian All-Ordinaries Index

This research study investigates the use of artificial neural networks to identify arbitrage opportunities in the Australian All-Ordinaries Index. Ten identically structured, independently trained, neural network committee members contribute their predictions on the Index movement. Trading decisions are made based on the unanimous consensus of their predictions and out-of-sample trading performance is assessed by the Sharpe Index. Empirical results show that technical trading based on neural network predictions outperforms the Buy-and-Hold strategy as well as “naive prediction”. Reliability of network predictions and hence trading performance is dramatically enhanced by the use of trading thresholds and the Committee approach