具有技术指标的逻辑回归模型预测谷歌股票的涨跌趋势  

Logistic Regression Model with Technical Indicators Predicts Ups and Downs For Google Stock Prices

作　　者：胡雪梅[1,2,3] 蒋慧凤 HU Xuemei;JIANG Huifeng(School of Mathematics and Statistics,Chongqing Technology and Business University,Chongqing 400067;Research Center for Economy of Upper Reaches of the Yangtse River,Chongqing 400067;Chongqing Key Laboratory of Social Economy and Applied Statistics,Chongqing 400067)

机构地区：[1]重庆工商大学数学与统计学院,重庆400067 [2]长江上游经济研究中心,重庆400067 [3]经济社会应用统计重庆市重点实验室,重庆400067

出　　处：《系统科学与数学》2021年第3期802-823,共22页Journal of Systems Science and Mathematical Sciences

基　　金：重庆市第五批高等学校优秀人才支持计划;重庆市科委基础研究与前沿探索一般项目(cstc.2018jcyjA2073);重庆市“统计学”研究生导师团队(yds183002);重庆市社会科学规划项目(2019BS061,2019WT59,2020YBTJ102);社会经济应用统计重庆市重点实验室平台开放项目(KFJJ2018066);重庆工商大学数理统计团队(ZDPTTD201906)和重庆工商大学高层次人才项目(1956003)资助课题。

摘　　要：正确预测股价的涨跌趋势会给投资者带来巨大的经济效益.近年来人们提出了时序预测、技术分析、基本分析和机器学习等方法来提高股价趋势的预测精度.文章主要结合技术指标和逻辑回归模型发展提高股价趋势预测精度的有效方法.首先,基于Murphy对金融股市发展的技术分析方法提取一些重要技术指标作为预测变量,构建刻画股价涨跌趋势的逻辑回归模型;再利用训练样本和迭代加权最小二乘法得到模型参数估计,计算股价上涨和下跌的概率估计,并确定最佳阈值预测股价涨跌趋势;最后利用检验样本计算混淆矩阵、灵敏度和特异度,绘制ROC(receiver operating characteristic)曲线评价预测精度.文章采用2010-2017年谷歌股价作为训练样本学习股价涨跌趋势,建立具有6个技术指标的逻辑回归预测2018年谷歌股价的涨跌趋势,不仅能得到股价涨跌的概率估计,而且能提高趋势预测精度和AUC(the Area under the ROC Curve).预测结果表明具有技术指标的逻辑回归预测方法优于支持向量机、人工神经网络、Elman神经网络和基于五类统计指标的一阶自回归逻辑模型.该方法也能预测美国微软等公司的股价涨跌趋势,给广大投资者带来更加丰厚的经济回报.Correctly predicting ups and downs for stock prices will bring huge economic profits for some investors.In recent years,some methods such as prediction of time series,technical analysis,fundamental analysis and machine learning techniques have been proposed to improve the prediction accuracy.In this paper we combine technical indicators with logistic regression model,and develop an efficient method to further improve the prediction accuracy for stock price trend movements.Firstly,we will base on technical analysis for finance stock market proposed by Murphy,choose some important technical indicators as predictors,and establish logistic regression model to describe ups and downs of stock prices.Secondly,we take advantage of the training set and the iteratively re-weighted least squares algorithm to obtain parameter estimators,compute the estimated probabilities for up and down trends,determine the optimal threshold values to predict up and down trends for stock prices.Finally,we apply the testing set to compute confusion matrix,sensitivity and specificity,and plot an ROC(receiver operating characteristic)curve to assess the prediction accuracy.In this paper we adopt Google prices from 2010 to 2017 as the training set to learn the law from up and down trends for Google prices,and establish logistic regression model with the six technical indicators to predict the two trends,not only obtain their estimated probabilities,for up and down trends,but also improve the prediction accuracy and AUC(the Area under the ROC Curve).The prediction results show that the proposed method performs better than support vector machine,artificial neural network,Elman neural network and the one-order auto-regression logistic model with the five basic statistic indexes.The proposed method can also predict up and down trends for other corporations such as Microsoft corporation,and bring the richer economic returns for the majority of investors.

关 键 词：技术分析 具有技术指标的逻辑回归 涨跌趋势 神经网络 支持向量机 

分 类 号：F831.51[经济管理—金融学]