This thesis introduces a new approach for the automatic detection of two crucially important shear wave splitting (SWS) parameters, fast wave polarization and delay time between split waves, from microearthquake seismograms. The method is based on the analyses of multiple time windows that include the shear wave arrivals. An automated SWS algorithm is performed for each specified window. Over the estimates of the two parameters (polarization and time delay) obtained from all windows, an unsupervised cluster analysis is applied to locate the region with the most stable estimate. The optimal region is that with the lowest variance. The mean value of the optimal cluster is regarded as the best estimate of polarization and time delay. The estimates are relatively easy to derive from large seismic datasets and show high reliability. We compare the results with manually estimated values of the SWS parameters from seismic data collected at The Geysers and Coso, CA, and Hengill, Iceland geothermal fields, and show that the method performs better than any other, providing up to 95% reliability (polarization) and 88% reliability (delay time) without human intervention.