Assessing Stock Market Time Series Predictors Quality through a Pairs Trading System

• Main Authors: Gomide, P, Milidiu, R L
• Format: Conference Proceeding
• Language: English
• Subjects: Artificial neural networks; Forecasting; Investments; Measurement; Pairs Trading; Predictive models; Stock Market; Stock markets; Trading System; Training
• ISSN: 1522-4899; 2375-0235
• DOI: 10.1109/SBRN.2010.31

The interest of both investors and researchers in stock market behaviour forecasting has increased throughout the recent years. Despite the wide number of publications examining this problem, accurately predicting future stock trends and developing business strategies capable of translating good predictions into profits are still great challenges. This is partly due to the nonlinearity and the noise shown by the stock market data source. And partly because benchmarking systems to assess the forecasting quality are not publicly available. Here, we present ANN models for both interday and intraday stock market forecasts. We also propose a trading system as a better way to assess the forecasting quality. The system is tested for Pairs Trading. We examine three pairs, composed by six assets of the top ten most traded companies of BM&FBOVESPA Stock Exchange, the world's third largest and official Brazilian stock exchange. The results are presented and compared to four benchmarks. The difference in the forecasting quality, when considering either the forecasting error metric or the trading system metrics, is remarkable. If we consider just the mean absolute percentage error, the ANN does not show a significant superiority. Nevertheless, when considering the trading system evaluation, it shows really outstanding results. The yields in some cases amount to a return on investment of more than 300%.