The eastern margin of the Pre-Caspian Basin is one of the most abundant structural zones for oil and gas accumulation。However, local carbonate rock reservoirs suffer from intricate diagenesis, widespread fractures, and high heterogeneity, complicating their complex pore structures and storage spaces. It is hard to identify and interpret logging response characteristics by only using one or two simple parameters. Focusing on the primary target layer KTII, a universal classification system is lacked with a dual-medium structure consisting of fractures and pores. This study quantitatively calculates the fracture development index in formations using the logarithmic difference between deep and shallow lateral resistivity curves, based on existing conventional logging, imaging, and core data in the study area. Integrating the fracture development index with conventional logging data, mechanistic models with data-driven machine learning models is innovatively merged to classify rock types into four classes. This approach avoids biases of single-criteria assessments, clearly distinguishes between porous and fractured reservoirs. With added permeability models and saturation height function, it greatly improves log interpretation and geological modeling precision. Applying this method to reevaluate previous logging interpretation results significantly improves the detection rate of thin reservoirs, aiding future exploration and development planning.