融合多种计算智能技术的股票价格时间序列预测建模研究

项目名称： 融合多种计算智能技术的股票价格时间序列预测建模研究

项目编号： No.71303067

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 管理科学

项目作者： 李嵩松

作者单位： 哈尔滨工业大学

项目金额： 22万元

中文摘要： 本项目将建立新的融合了多种计算智能技术和隐马尔可夫模型的针对于股票价格时间序列的混合预测模型。通过将计算智能技术中的几个核心方法：人工神经网络ANN，遗传算法GA，模糊逻辑FL，和进化算法EA，逐层递进地加入到基于隐马尔可夫模型所构建的基础预测模型当中，从而改进和完善该预测模型，最终得到拥有较高预测精度并同时也拥有较高预测效率的混合预测模型。本项目将致力于解决以下科学问题：（1）建立新的针对于股票价格时间序列的混合预测模型，解决目前研究中缺少将多种技术混合的预测方法的问题；（2）通过引入计算智能技术中的FL方法，处理股票价格时间序列数据的非线性且非平稳性，提高混合预测模型的预测精度；（3）通过引入计算智能技术中的EA方法，解决混合预测模型在提高预测精度的同时产生大量模糊规则的问题，在最优模糊规则数目和预测准确性之间，找到折中解的范围。最后，通过实证和比较研究，对新混合模型进行检验。

中文关键词： 股票价格时间序列；计算智能；模糊逻辑；进化算法；隐马尔可夫模型

英文摘要： This project will establish a new hybrid forecasting model combined with a variety of computational intelligence technologies and hidden Markov model to predict the stock price time series. This project will put a few core methods of computational intelligence technologies: artificial neural network (ANN), genetic algorithm (GA), fuzzy logic (FL) and evolutionary algorithm (EA), stepwise progression into the basic forecasting model which is built based on hidden Markov model (HMM), to establish the perfect hybrid forecasting model, so as to further improve the prediction accuracy at the same time increase the prediction efficiency. This project will be devoted to solve the following scientific problems: (1) establish a new hybrid forecasting model to fill in the blank of present research, which lacks the hybrid prediction method combining a variety of technologies; (2) deal with the nonlinear and non-stationary of the stock price time series by computational intelligence technology, FL, to improve the prediction accuracy; (3) solve the problem of produce a large number of fuzzy rules when improve the prediction accuracy by computational intelligence technology, EA, and find the range of compromise solutions between the optimal number of fuzzy rules and the prediction accuracy. Finally, through the empirical and

英文关键词： stock price time series；computational intelligence；fuzzy logic；evolutionary algorithm；Hidden Markov Model