Rock classification or assigning a type or class to a specific rock sample based on petrophysical characteristics is a fundamental technique to reduce the uncertainty in prediction of reservoir properties due to heterogeneity. One of the popular approaches is to classify a reservoir rock from the fundamentals of geology and the physics of flow at pore network scale. In this approach, rocks of similar fluid conductivity are identified and grouped into classes. Each grouping is referred to as a hydraulic flow unit (HU). HUs are estimated for the uncored section of the wells from wire line log data received from field using three popular artificial intelligence (AI) techniques such as (i) Adaptive Network Fuzzy Inference Systems (ANFIS); (ii) Functional Networks; and (iii) Support Vector Machines. HUs can either be predicted directly from wire line log data as a classification model or can be estimated indirectly from predicted permeability and porosity. In the present study, HUs were predicted employing both approaches. The comparison of the two approaches shows that it is a better practice to estimate HUs by directly relating them to well logs. Calculation of HUs from predicted permeability and porosity results in inferior accuracy. This is due to the accumulation of errors during the estimation of permeability and porosity. The Knowledge of HUs acquired from well logs can significantly help in determining ultimate hydrocarbon recovery, optimal well placement, and well stimulation. Thus, the finding of this research will impact the economy of the development and operations of a field.