Stock medium and long term trend prediction method and system based on Bayes classifier

Abstract

The invention relates to a stock medium and long term trend prediction method based on a Bayes classifier. The method comprises the steps of selecting stock data, and determining all starting points and an interval length dj; dividing a compartment, and calculating the interval slope of historical data; learning the interval slope of the historical data and predicting confidence coefficient judgment compartments, so as to obtain the average price of stocks of a plurality of trading days by taking the confidence coefficient judgment compartment as the starting points; calculating confidence, and comparing the confidence and a preset threshold value; predicting a future compartment slope, and converting the future compartment slope to obtain the average price of the stock of the plurality of trading days by taking the prediction interval starting points as the starting points; normalizing the ups and downs of the average price of the stock of the plurality of trading days by taking the prediction interval starting points as the starting points; and building a stock tank. According to the method provided by the invention, the accumulative errors can be prevented, the trend change of the stocks in the prediction intervals can be displayed, the fluctuating change tendency of a stock market can be caught better, and the transaction exposure can be effectively estimated.

Description

technical field [0001] The present invention relates to a data processing system or method specially suitable for administration, commerce, finance, management, supervision or forecasting purposes, especially a medium and long-term stock trend forecasting method and system based on a Bayesian classifier. Background technique [0002] The stock market is an important place for the optimal allocation of capital resources, and mastering its changing laws is not only a dream for investors, but also has important practical significance for the research and management of the macro national economy. Since the factors affecting the stock price include internal factors of the enterprise, economic factors, system factors, people's psychological factors, etc., and the degree and mode of influence of various factors are different, it is very difficult to accurately predict the stock price. [0003] Time series analysis has become an indispensable part of stock market research both theor...

AI Technical Summary

Problems solved by technology

[0004] At present, the medium and long-term forecasting methods of existing stocks mainly include: using the forecast results of the previous step to iteratively and recursively obtain the subsequent medium and long-term forecast (reference example: Yang Yiwen, Lin Yupei. Fuzzy time series modeling and stock market multi-step Forecasting [J]. Computer Engineering and Application, 2014, (5): 252-256.), but there is a cumulative error in this method, and the cumulative error increases with the growth of the forecast step; using the moving window algorithm in the modeling sequence The forecasting model is recursively updated by deleting some old data and incorporating some new data (reference example: Jian Qingming, Zeng Huanglin, Ye Xiaotong. Application of support vector regression based on moving window and dynamic optimization in index forecasting[J].Computer application and Software, 2011, (12): 83-85.), the length of the moving window in this method has a great influence on the modeling accuracy, and it can only show the average stock price in the prediction interval, but cannot show the stock trend in the prediction interval changes in details

In addition, the current methods directly use the mean value as the characteristic value of stock trend prediction. For time series with large fluctuations, the mean value will weaken the fluctuation characteristics of the time period and reduce the accuracy of medium and long-term forecasting.

Images