This article presents the implementation of the sparse twin least squares support vector regression (WTLSSVR) model, which is derived from the twin least squares support vector regression (TLSSVR) model. The WTLSSVR model achieves sparsity by eliminating sample points with small Lagrange multipliers.Our algorithm involves combining the LSSVR with the twin hyperplane algorithm, removing sample points with small Lagrange multipliers, and thinning the model to construct a sparse twin least squares support vector regression model. This approach enhances the predictive accuracy of current twin regressors, while maintaining excellent generalization performance.Therefore, the WTLSSVR is superior to the traditional LSSVR in generalization performance. In addition, compared with some other traditional algorithms, WTLSSVR has higher prediction accuracy, which is obvious when dealing with large-scale problems. To validate the proposed model’s soundness and practicality, we carried out numerous experiments, encompassing two UCI datasets and a genuine vegetable price dataset. The results of the experiments indicate that the suggested algorithm is efficacious, practical, and versatile.