Trend predictions of tick-wise stock prices by means of technical indicators selected by genetic algorithm

We propose a systematic method for predicting the trend of the price time-series at several ticks ahead of the current price by means of a genetic algorithm, used to optimize the combination of the frequently used technical indicators such as various moving averages, the deviation indicator from the moving averages, and so on. We show that the proposed method gives good predictions on the directions of motion, with the rate as high as 80% for multiple stocks of NYSE selected from four different business types. We also show that the performance improves if we combine two or three indicators compared to the case of using a single indicator. However, the performance seems to go down as we increase the number of the indicators from the optimum value. This work was presented in part at the 12th International Symposium on Artificial Life and Robotics, Oita, Japan, January 25–27, 2007