Prediction of probabilistic seismic hazard inclusive of local site effects across Metro Vancouver is a multi-faceted challenge involving multi-source seismicity and the greatest extremes in geologic sub/surface complexity within Canada. To achieve probabilistic shaking amplification hazard mapping inclusive of 1D site and 3D sedimentary basin effects for western Metro Vancouver, a comprehensive database of seismic site condition measures at over 15,000 locations was compiled from existing public and private data sources in combination with performing 2,787 single-seismometer amplification surveys, 180 seven-seismometer surface wave dispersion surveys, and a regional-scale nineteen-seismometer ambient noise tomography survey. Regional seismic hazard mapping was generated efficiently from three key (of seven trialed) shaking susceptibility maps (providing the microzonation) based on ground motions predicted from 1D SRAs performed using selected earthquake waveforms scaled to the uniform hazard spectrum of each seismic source at the two chosen return periods to provide site amplification factors at 6 selected spectral periods (the amplification values for mapping). At long spectral periods (≥ 2 seconds), ground motions are predicted from physics-based 3D wave propagation simulations of large-magnitude scenario earthquake ruptures within regional 3D velocity models (60 km depth) to provide basin-only amplification factors where the average basin depth is greater than the average basin depth for a given Vs30. Such a physics-based data-driven methodology is accurate but cumbersome. As we progress with seismic microzonation mapping of eastern Metro Vancouver, we are applying regional data-pattern-recognition and machine learning approaches to improve efficiency and accuracy(?) of probabilistic seismic hazard prediction at regional spatial scales.