Screw conveyors are widely used in granular transportation to provide an efficient and steady flow rate. DEM is a numerical method used to predict flow behaviors of granular material effectively. However, this method is computationally intensive. In this work, an artificial network model was trained using DEM simulation results to reduce computational cost while keeping the estimation accuracy. The main drawback of this technique is that it requires large number of data which is time consuming when a series of DEM simulation results with varying parameters are used to train the network mainly for screw conveyor applications. To get beyond this limitation, the DOE approach was used to optimize ANN parameters by performing significantly fewer virtual experiments. Moreover, the effect of particle shape on mass flow rate was also considered using single and clumped spheres. The trained artificial neural network was able to predict mass flow rate accurately and highly efficiently by taking into account both screw conveyor related parameters and granular particle related parameters as inputs. For validation of ANN, experimental tests were performed using polypropylene granular material. The findings showed that the proposed model by ANN was also in a good agreement with the experimental data for horizontal screw conveyor.