The ground shaking hazard at a particular site can be estimated quantitatively using probabilistic or deterministic approach. The probabilistic way of analyzing seismic hazard is reasonable in considering the seismicity distribution of all possible magnitudes and all possible distances over a specified time period. The conventional way of defining the seismicity by dividing the seismic source into zones for distributed seismicity, characterizing each zone with Gutenberg-Richter recurrence relationship and assuming uniform seismicity within each zone has certain drawbacks. In this study, zone free approach is proposed to evaluate the spatial distribution of seismicity based on non-parametric density estimation technique i.e. kernel density estimation technique (KDE). The fixed bandwidth kernel deals poorly with the earthquake distributions since the earthquake catalogue have several areas of high activity clusters and low background seismicity. Therefore in the present study, adaptive kernel technique using clustering method is proposed and applied to find the spatial activity rate density functions and integrated with other forms of uncertainty in magnitude and distance to determine the annual rate of exceedance of the selected ground motion parameter. In order to test the practical applicability of the proposed method, Chennai region (low to moderate seismicity region) has been chosen which lies in the intraplate seismicity region with diffused and distributed seismic characteristics. The results of the hazard analysis are provided in the form of peak ground acceleration (PGA) and uniform hazard spectra (UHS) for different return periods at the bed rock level. The PGA and UHS obtained from the present study for Chennai are compared with the Cornell-McGuire and the fixed kernel methods.