Predicting the tick-wise price fluctuations by means of evolutionary computation

Tick-wise price fluctuations are known to have a certain characteristic style unlike the daily-close prices whose increments can be approximated by the random walk. We then expect to make a tick-wise prediction by utilizing those characteristic features. We apply an evolutional algorithm to automatically generate a set of predictions that suit the changing environment given by price fluctuation data. Based only on past price fluctuations, we have obtained up to a 70% hit rate for the direction of the next motion by using 1-6 ticks in the immediate past.