In this study, rough elliptic bore journal bearing performance is predicted using an artificial neural network (ANN) technique. The effects of non-circularity and roughness are quantified to elliptic and isotropic in macro and micro scale, respectively. The numerically estimated performance parameters like load, friction, and flow-in at different eccentricities [0.3 (low), 0.5 (medium), and 0.8 (high)], non-circularities [0.5 (low), 1.0 (medium), and 2.0 (high)], and roughness factors [0.1 (low), 0.2 (medium), 0.3 (medium), and 0.4 (high)] are used to train and build the ANN model. The training continued until the maximum mean square error is achieved, and the best-fitting plot is generated. With a confidence level of 99.75% or an R-value of 0.99757, the results predicted are found to be satisfactory.