Research Article
Stock Return Prediction Using Machine Learning Regression Models
@INPROCEEDINGS{10.4108/eai.22-5-2026.2365163, author={Ning Deng}, title={Stock Return Prediction Using Machine Learning Regression Models}, proceedings={Proceedings of the 4th International Conference on Image, Algorithms, and Artificial Intelligence, ICIAAI 2026, 22-24 May 2026, Singapore, Singapore}, publisher={EAI}, proceedings_a={ICIAAI}, year={2026}, month={8}, keywords={Short-term stock return prediction; Machine learning regression; Financial time series; Model comparison; Historical price data}, doi={10.4108/eai.22-5-2026.2365163} }- Ning Deng
Year: 2026
Stock Return Prediction Using Machine Learning Regression Models
ICIAAI
EAI
DOI: 10.4108/eai.22-5-2026.2365163
Abstract
This study analyses the hypothesis of whether machine learning regression models can forecast the stock returns of stocks in the short term on the basis of historical market data. An integrated pipeline of time-series is run to facilitate uniform preprocessing, construction of features and evaluation of models. The dataset has 615, 505 observations that can be used representing 505 U. S. stocks between 2013 and 2018. The input features are lagged daily returns over the last five trading days to forecast trading day returns. Experimental results indicate that linear regression has an MSE of 2.17 × 10⁻⁴. Decision tree regression achieves an MSE of 2.19 × 10⁻⁴. The neural network model performs the worst among the three, with an MSE of 3.07 × 10⁻⁴. These results suggest that prediction of short-term returns is very difficult and that more complex model does not necessarily enhance better predictive accuracy.
