Seismic processes exhibit complicated and nonlinear behavior, making earthquake forecasting difficult. Earthquake prediction is framed in this study as a regression problem. It uses laboratory acoustic emission signals from the LANL Earthquake Prediction dataset to predict continuous time-tofailure (TTF). Support Vector Regression (SVR), Kernel Ridge Regression (KRR), and Light Gradient Boosting Machine (LightGBM) are the three regression models used to analyze statistical information taken from segmented acoustic windows. To provide fair comparison, all models are assessed using the same preprocessing procedures and 10-fold cross-validation. SVR has the lowest mean absolute error, according to the results, even if there aren't many variations between the models. Overall, a systematic comparison of regression techniques for controlled seismic forecasting scenarios is presented in this article.