Local site effects significantly influence seismic ground response analysis, playing a crucial role in the seismic design of structures. Ground response analysis, which predicts response of soil deposits to various earthquake motions, employs three different approaches: linear, equivalent linear, and nonlinear methods. Among these, the equivalent linear method is widely preferred for its capacity to simulate nonlinearity in soil behaviour and its time efficiency. This study introduces an equivalent linear site amplification model that utilizes a linear regression approach in machine learning to predict site amplification. The developed model incorporates six input parameters: thickness of the soil column (H), shear wave velocity of the soil column (VS), shear wave velocity of the elastic half-space (VHS), peak ground acceleration (PGA) of the input motion, duration (T) of the input motion, and predominant frequency (f) of the input motion. The model outputs the site amplification factor. To train and test the machine learning model, data obtained from DEEPSOIL, a 1D ground response analysis software, were utilized. The training set comprised 80% of the DEEPSOIL datasets, while the remaining 20% were used for testing. Model performance was evaluated by comparing the amplification results from the model with those from DEEPSOIL. A favourable agreement between the two was observed. The comparison of results obtained from the linear regression model and DEEPSOIL was visualized using a kernel density plot, revealing a consistent match between the model and the DEEPSOIL software. This study demonstrates the efficacy of the linear regression approach in predicting site amplification, offering a valuable tool for seismic ground response analysis in structural design.