Abstract: Due to the inconsistent tendency of stock price and technical index,the share price trend prediction algorithm based on the technical features performs poorly.A share price trend prediction algorithm from the point of deviated characteristics(DCPA) was proposed.Deviated features firstly are extracted, the degree of swerve is calculated,and then the trend of price is forecasted by BP neural network according to the degree of swerve and the stock’s closing price.While the risk appetite is high,the correlation between stock price trend and deviated characteristics is weak,thus a risk preference based share price trend prediction algorithm named RPDCA was put forward on the basis of DCPA algorithm.Firstly, features which are associated with risk appetite are extracted and the current market risk preference type is acquired through the risk appetite computational model.Secondly,by means of Bayesian network,the structural relationship among risk appetite,deviated characteristics and the trend of stock price is learned,and then the interdependent relationship between risk appetite and deviated characteristics is analyzed by using node asymmetric information entropy.Last,the trend of stock price is forecasted self-adaptedly according to the relationship between risk appetite and deviated characteristics by BP neural network.Based on the comparison and analysis on the actual data,the experimental results show that the RPDCA algorithm has higher precision of short-term prediction.

Key words: Deviated characteristic,Risk preference,DCPA algorithm,Utility function,RPDCA algorithm