基于形态特征与因果岭回归的股市态势预测算法  

Stock Market Trend Prediction Algorithm Based on Morphological Characteristics and Causal Ridge Regression

作　　者：姚宏亮[1] 马晓琴[1] 王浩[1] 李俊照[1] 

机构地区：[1]合肥工业大学计算机与信息学院,合肥230009

出　　处：《计算机工程》2016年第2期175-183,共9页Computer Engineering

基　　金：国家自然科学基金资助项目(61175051;61070131;61175033)

摘　　要：基于股票波动典型的M形态,提出一种基于因果关系的岭回归股市态势预测算法。根据M形态的波动特征,引入能量思想,以M形态的边、波峰和波谷为结点,构建M形态的贝叶斯网络结构模型。利用马尔科夫毯算法和非对称信息熵,得到M形态的局部因果结构。采用因果强度的度量标准,将M形态因果关系引入到岭回归模型中,对股市态势进行预测。该模型通过将股票形成和能量波动的因果关系相结合,可以有效地发现股市的突变点。真实数据集上的实验结果表明,相比标准的岭回归算法和基于径向基的神经网络算法,该算法具有更好的预测效果。Based on the typical M form of the block volatility,this paper puts forward a ridge regression stock market trend prediction algorithm based on causality. Stock form reflects the stock fluctuations of energy change. According to the characteristics of the fluctuations in the form of M introducing energy ideas,based on edge,peaks and troughs in the form of M nodes,it builds a Bayesian netw ork structure model in the form of M. By using Markov blanket algorithm and asymmetric information entropy,it gets a local causal structure in the form of M. The introduction of the strength of causal metrics is introduced to the M shape causality in ridge regression model for its stock market trend prediction of the model through stock form and causation of energy fluctuations,which can effectively find the abrupt change point of the stock market. Results on real data sets show that,compared with ridge regression algorithm and radial basis neural netw ork algorithm,the proposed algorithm has better prediction effect.

关 键 词：贝叶斯网络 能量模型 因果分析 岭回归模型 预测算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]