Equity trend prediction with neural networks

Abstract

This paper presents results of neural network based trend prediction for equity markets. Raw equity exchange data is pre-processed before being fed into a series of neural networks. The use of Self Organising Maps (SOM) is investigated as a data classification method to limit neural network inputs and training data requirements. The resulting primary simulation is a neural network that can prediction whether the next trading period will be, on average, higher or lower than the current. Combinations of pre-processing and feature extracting SOM’s are investigated to determine the more optimal system configuration.