Low fidelity modeling approaches remain attractive due to an unrivaled ability to predict full turbine performance maps quickly compared to high-fidelity approaches such as computational fluid dynamics (CFD), especially in the preliminary design process. As improvements in performance on a component level approach a point of diminishing returns, the ability to efficiently optimize the complete charging system for a given duty is a topic attracting significant research interest. In the case of turbocharging applications, existing engine and powertrain simulations require turbine maps to calculate the turbine performance, which are usually obtained from experimental testing. Unfortunately, the need for extrapolation is unavoidable because of the limited range of testing data available, leading to inaccuracies especially at off-design conditions. To enable intensive modeling and optimization of complete vehicle powertrains for different drive cycles, the current piece of work seeks to combine the advantages of machine learning techniques and physical meanline modeling to facilitate faster, more accurate predictions of complete turbocharger maps. This paper presents a novel methodology for turbocharger turbine rotor and nozzle performance prediction based on hybrid modeling. The turbine rotor and nozzle were parameterized to conduct CFD simulations for a wide variety of turbine geometries, which were used to form a database to train an artificial neural network (ANN). The predicted losses provided by the ANN were then utilized in the meanline code, substituting for the conventional empirical loss models. As well as removing the need for empirical loss models, modifications were undertaken to the meanline approach to further enhance modeling accuracy. First, in order to accurately characterize the stage mass flow capacity, the losses occurring in the nozzle and rotor were subdivided into those occurring before and after the throat. A second novel aspect is that the aerodynamic blockage level at the rotor throat was implemented as a variable rather than a constant value. By training the ANN to predict the variation of blockage with geometry and operating condition, a more accurate depiction of the changing secondary flow fields could be achieved. The capabilities of the hybrid meanline modeling method were evaluated on several unseen test cases. The resulting predictions of efficiency and mass flowrate demonstrated strong correlation with CFD results and experimental test results. The hybrid meanline modeling method therefore displays great potential in wide range radial turbine performance prediction with enhanced accuracy in comparison to traditional approaches.