Analyze Long Mid-term Trends of Stock with Genetic Programming on Moving Average and Turning Points

摘要

This paper employs Genetic Programming (GP) with individuals of tree structure to form empirical formulas in order to track the dynamic pattern of the moving average curves of stock prices. We find that our method tracks the 60-day moving average better than other shorter period averages. In order to minimize the effects of noise and other random events impacting on the markets and maximize the effective information abstracted from the origin data, two comparable data preprocessing methods for turning points are proposed to cooperate with GP for more stable long & mid-term dynamic analysis and prediction. We use either discrete data with fixed time intervals as long as 120 days or data at local extreme by FFT. So, the formula finding system tracks the next turning point with the information of several previous turning points. Simulations show that our method to track and predict long & mid-term change trend of stock price is practical.