11. Multi-variate financial index prediction using neural networks [138] (1990), with Tony Harker

C G Windsor and A H Harker Multivariate Financial Index Prediction - A neural network study, INNS Conference 1990 Klewer Academic Publishers, 357-360, 1990.

Always aware of the big application, Windsor was one of the first people to try out neural network methods for the prediction of the stock market. The idea is to learn from previous experience by getting the neural network to assimilate the fluctuations of the market over a series of previous years, and so detect a similar pattern if it occurs again. There had been a few previous studies of this problem, but Windsor's was the first that used multiple indices to predict a chosen index. Thus interest rates, money supply, and even the Conservative party majority were included in the training data used to predict the FT index. The paper and poster attracted enormous interest at the Paris International Neural Network Conference in 1990. Many banks and financial institutions are now working in this field although naturally the real results are not published.