Due to intrinsic complexity under mixed traffic conditions, the car-following as well as lane-change models developed for homogeneous traffic conditions are not directly applicable, as there are very frequent longitudinal and transverse movement of different vehicle classes, simultaneously on same available road space with weak lane-discipline. Moreover, under mixed traffic stream, there can be multiple combinations of following and leading vehicle pairs over space and time. On these lines contingent heavily on the computational driving behavior models in coding the autonomous vehicle may not be productive, considering this in the present work supervised machine-learning models, which is a part of artificial intelligence are employed in modelling the driving behavior. For this purpose, vehicular trajectory data, which is minute level data source where vehicles instincts will be recorded over time and falls under big data stream for vehicular behavior is employed. The performance of the algorithms using trajectory data was evaluated among different criteria. From that k-NN algorithm is found to be the most suitable candidate in replicating the driving behavior accurately under heterogeneous traffic.