Earthquakes are random triggering phenomena that produce spatial and temporal clusters, thus creates a bias in a seismic catalogue. Seismicity de-clustering is the process to identify mainshocks, aftershocks-foreshocks and backgrounds in a historical catalogue of an earthquake prone regions. The seismicity de-clustering finds extensive applications in design of earthquake prediction models and seismic hazard assessment. The multi-objective optimization algorithms are preferred over their single objective counter-parts as they offer the end user the flexibility to select any solution from a set of optimal solution in the form of Pareto Front. Two objective functions : Coefficient of variance (COV) in temporal domain, and m-Morisita Index (m-MI) in spatial domain are formulated for simultaneous optimization for seismicity de-clustering. Popular nature inspired multi-objective algorithm Nondominated Sorted Genetic Algorithm (NSGA-II) is employed for the optimization task. The reported model are tested on thirty two years earthquake catalogue of Himalaya, Japan and California region. Comparative analysis reveal accurate determination of aftershock and background events over other benchmark statistical models like : Gardner and Knopoff (GK) model, Reasenberg model and recently developed tetra-stage cluster identification model. The results are decimated in the form of cummulative plots, inter-event time versus inter-event distance plots, and Lambda plots.