We present a data-driven model for tectonic zonation of the West Siberian Basin (WSB) based on K-means clustering applied to a multivariate geophysical dataset. The initial analysis incorporates lithospheric thickness, crustal thickness, sedimentary thickness, topography, surface heat flow, and S-wave velocity anomaly at 100 km depth. F-statistics and permutation feature importance analysis indicate that only four primary parameters are sufficient to achieve reliable zonation, yielding six clusters that correspond to distinct tectonic domains. These domains reflect the complex geodynamic evolution of the region, including Paleozoic accretion, Mesozoic rifting, and subsequent subsidence. Independent data on hydrocarbon field locations demonstrate that major oil accumulations are primarily associated with two of these clusters, supporting the validity of the approach. The resulting zonation provides a reproducible basis for lithospheric regionalization and resource assessment. This framework can be further developed by integrating supervised machine learning methods to predict a specific structure and thermal regimes in areas with limited data. The quantitative characterization of these domains provides an objective framework for future geodynamic models and resource assessments.