Scholarlink

<p>Accurately predicting weekly stock up or down trends is critical in the financial markets. In order to improve the prediction accuracy, in this paper we combine technical indicators, fundamental indicators and sentiment indicators with LASSO/SCAD/MCP penalized logistic regressions for predicting weekly up or down trends of stock prices for listed companies. Firstly, 24 technical indicators, 20 fundamental indicators and 4 sentiment indicators are selected as predictor vectors, and weekly up or down trends of stock prices is used as a two-class response variable, and then construct LASSO/SCAD/MCP penalized logistic regressions. Secondly, coordinate descent algorithm and the training samples are introduced to select some important variables and get these parameter estimators, and apply the test samples and the trained models to predict weekly up or down trends of stock price. Thirdly, we calculate the confusion matrix, sensitivity and specificity, and plot the ROC curve to compare their prediction performance. Finally, we apply the proposed method to predict weekly up or down trends for four Chinese listed companies from 2010 to 2023. The prediction results show that LASSO/SCAD/MCP penalized logistic regressions with technical indicators, fundamental indicators and sentiment indicators outperform LASSO/SCAD/MCP penalized logistic regressions with technical indicators, 3 machine learning methods and 3 deep learning methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Penalized Logistic Regressions with Technical Indicators, Fundamental Indicators and Sentiment Indicators Predict Weekly Stock Up or Down Trends

  • 期刊论文

  • 发表日期：

    2026年1月14日

查看全文
  • Xuemei Hu,
  • Liping Liu,
  • Tingqiong Cui,
  • Lianxing Xia

摘要

Accurately predicting weekly stock up or down trends is critical in the financial markets. In order to improve the prediction accuracy, in this paper we combine technical indicators, fundamental indicators and sentiment indicators with LASSO/SCAD/MCP penalized logistic regressions for predicting weekly up or down trends of stock prices for listed companies. Firstly, 24 technical indicators, 20 fundamental indicators and 4 sentiment indicators are selected as predictor vectors, and weekly up or down trends of stock prices is used as a two-class response variable, and then construct LASSO/SCAD/MCP penalized logistic regressions. Secondly, coordinate descent algorithm and the training samples are introduced to select some important variables and get these parameter estimators, and apply the test samples and the trained models to predict weekly up or down trends of stock price. Thirdly, we calculate the confusion matrix, sensitivity and specificity, and plot the ROC curve to compare their prediction performance. Finally, we apply the proposed method to predict weekly up or down trends for four Chinese listed companies from 2010 to 2023. The prediction results show that LASSO/SCAD/MCP penalized logistic regressions with technical indicators, fundamental indicators and sentiment indicators outperform LASSO/SCAD/MCP penalized logistic regressions with technical indicators, 3 machine learning methods and 3 deep learning methods.