The precise prediction of sky luminance distribution is inevitable for design of efficient daylighting systems. The lack of measured sky luminance data in developing countries like India makes the current standard design sky models redundant. Most of the designer use Perez sky model for prediction of sky luminance distribution using solar insolation conditions for such locations. Data-driven models using machine learning can be a possible solution for prediction of sky luminance distribution in such locations. The present study compares the prediction accuracy of such data-driven models using four different machine learning techniques and compares the results Perez sky model. It becomes clear from the results that the data-driven model does not perform well when the location is different from the training dataset location, and the Perez sky model gives better results for such cases.