Comparing time series and associative classification approaches to quantitative stock trading

Comparing time series and associative classification approaches to quantitative stock trading. Can time dimension become negligible in time series forecasting? This work tries to answer that question in a specific, still interesting, domain: economic quantitative analysis. The study compares a new, time-independent methodology, built upon data mining and association rules extraction, to a set of classical, time-dependent approaches for time series analysis. Forecasting methods have been used to build a trading system that impersonates a daily trader operating on Italian stock exchange market. Models performances have been compared in term of profit and accuracy. Results show that time-independent approach is comparable to time-dependent methods.