The Research Of Stock Market Trend Forecasting Based On Peer Group

Stock market trend forecasting for investment decisions has an important guiding role. However, stock movements are affected with macroeconomic policies, incidents and manipulated in stock market and other factors, so difficult to predict stock movements. In this dissertation, for the mutability and variability problem, SFM-PG algorithm and TM-PG algorithm are proposed, to solve the stock of short-term and medium-term trend forecasting problems, specific research work is as follows:(1) Analyze the feasibility of combined with Markov blanket models and peer group analysis method. The traditional algorithm uses peer group average weight method for establishing peer groups target tracking model, the dissertation established a target object tracking model using the weighted average method, enabling peer group members more closely track their targets.(2) SFM-PG algorithm is proposed in this dissertation. SFM-PG algorithm to select the target stock Markov blanket as their peer groups, peer groups based on proximity between the tracking window gives a prediction model, by the weight of the peer group dynamic updates tracking forecasts, reducing stock data distribution on non-normality of prediction; Furthermore, the use of a sliding window to extract the characteristic of time series data and the formation of flow characteristics, feature extraction flow pattern matching and pattern through the knowledge base, and use the knowledge flow characteristic pattern corresponding adjustment predicted to reduce the prediction error introduced mutations. The results permit the needle above the stock industry sector show that the algorithm has high prediction accuracy.(3) In the SFM-PG algorithm to obtain empirical analysis of the practicality and effectiveness, the TM-PG algorithm is proposed combined with peer group algorithms and behavioral transmission, the stock index movements and behavior problems among transmission were studied for predicting the stock offers another idea. Stock index for the empirical analysis, the results show that in the medium term, this algorithm has better prediction precision, at the same time to verify the changes in trends between composite has a transmission.