Perovskites are a unique class of semiconductors with long carrier diffusion lengths which have shown remarkable promise for photovoltaic applications as their solar conversion efficiencies increased rapidly by nearly 20% over the past decade. Here we evaluate the efficacy of artificial neural networks for the analysis and prediction of 3D perovskite band gaps; which is the key parameter used in solar cell design. The set of input parameters for our artificial neural network includes descriptors such as s, p, d, and f orbital radii, electronegativity, octahedral factor, and formation energy. Furthermore, we apply a recursive feature correlation filter to eliminate features which highly correlate with other features in our model as well as features which have a low contribution with regards with regards to predicting bandgap. We find that splitting the dataset between training, testing, and validation sets in a 70:15:15 ratio provides a highly reliable model that yields a mean square error of less than 0.25 eV. We then validate the capability of artificial neural network by testing its predictability for bandgaps based on unseen inputs.