This study investigates the petroleum reservoir potential of the Palaeocene Lockhart, Hangu, and Patala formations in the Kohat Basin, Pakistan, by integrating petrophysical analysis and unsupervised K-means electrofacies clustering (Zhang and Postaire, 1994) on well log data from Makori-01, Manzalai-02, and Sumari Deep X-01. Key reservoir parameters, including porosity (12-18%), shale volume (Vsh < 20%), and water saturation (Sw < 40%), were derived using density-neutron crossplots, gamma-ray log interpretations, and Archie's equation (Archie, 1942). The Lockhart Formation exhibited the most favorable reservoir quality, with 15% of its intervals classified as high-porosity (14-18%), low-Sw (<35%) electrofacies (EF2, EF4), correlating to fractured limestone. The Hangu Formation showed heterogeneous sandstone reservoirs, where 23% of intervals (EF13, EF15) demonstrated moderate porosity (12-16%) and Sw < 40%, aligning with fluvial-deltaic depositional models (Ahmed et al., 2014; Jan et al., 2019). The Patala Formation, though predominantly shaly (EF1, EF3), revealed localized sandstone electrofacies (EF6) with porosity ~12%. Cross-validation confirmed strong alignment (R² = 0.78) between electrofacies clusters and petrophysical "sweet spots," with 85% consistency to stratigraphic markers (Ghani et al., 2022). These results resolve vertical heterogeneity and identify untapped reservoir zones, challenging prior dismissals of the Patala Formation (Ali et al., 2022). The workflow advances reservoir characterization in structurally complex basins, offering a 30% improvement in lithological discrimination over conventional methods, as demonstrated in Himalayan foreland analogs. This study underscores the Kohat Basin's potential for unconventional hydrocarbon exploitation, particularly in the Lockhart and Hangu formations, and advocates for machine learning-driven approaches in frontier exploration.