Logic Mining for Financial Data

In this paper we consider a new strategy for supporting timing decisions in stock markets. The approach uses the logic data miner Lsquare, based on logic optimisation techniques. We adopt a novel concept of good session, based on the best return expected within a given time horizon. Such definition links indirectly the buying decision with the selling decision and make it possible to exploit particular features of stock time series. The method is translated into an investment strategy and then it is compared with the standard buy & hold strategy.