Identifying Stock Similarity Based on Episode Distances

Predicting stock market movements is always difficult. Investors try to guess a stock´s behavior, but it often backfires. Thumb rules and intuition seems to be the major tools. One approach suggested that instead of trying to predict one particular stock´s movement with respect to the whole market, it may be easier to predict a stock A´s movement based on another stock B´s movement, because A may get affected by B after B´s movement. This may provide the investor invaluable time advantage. It would be very useful if a general framework can be introduced that can predict such dependence between stocks based on any user defined criterion. This article attempts to lay down one such framework, where the stock time series is encoded as a binary string. This binary representation depends on the user defined criterion. The string distances between two such encoded time series has been used as a measure of dependence. Further, this technique has been used in the dasiapairs trading strategypsila; in fact, it is more powerful as varied user defined criterion can be handled while detecting similarity. The presented technique has been demonstrated with one typical user defined criterion.