Effects of wake from turbines significantly influencing the power production in a wind farm are crucial part in wind farm design. Several wake models starting from the analytical to the Computational Fluid Dynamics (CFD) are developed in the literature. The analytical models are simple but not accurate, whereas large eddy simulations (LES) models are accurate but time-consuming. The major concern of analytical models is that they consider linear wake growth rate which might not be true. To address this issue, a data-driven wake model under the machine learning framework is proposed in this work to reconcile the limitations of analytical wake models. Further, the data-driven wake model is compared with the power produced through analytical models provided in the literature such as Jensen model to validate its efficiency. The proposed model has shown better accuracy when compared with other models, justifying the significance of nonlinear growth rate of wake and its impact on power production through wind farm.