Data-driven approaches have the potential to make a priori predictions. However, there are very few models that have been explored for the prediction of nanocomposite electrocatalysts under testing conditions. Here we report for the first time the parametric optimization coupled with data-driven approaches of efficient electrocatalysts for the oxygen evolution reaction (OER) in alkaline media. The parametric optimization suggests that the porous Ni- or Fe-based nitride composites, with combinatorial structures or thin films with average d-electrons between 5 and 8 and catalyst loading between 1 and 10 mg cm−2 will exhibit the best OER activity. Machine learning classification of OER overpotential grades (η10, overpotential at 10 mA cm−2) of transition metal oxynitrides (TMONs) and transition metal oxides (TMOs) is achieved using random forest (RF) and CatBoost algorithms. η10 is classified as grade ‘A’ if η10 < 351 mV, else grade ‘F’. We note that 80–85% of electrocatalysts containing nickel foam (NF) have been in grade A, implying NF is a prospective hindrance against true activity determination of the electrocatalyst but suitable for achieving grade ‘A’ electrocatalyst for electrolysers. RF and CatBoost models achieved an accuracy of 78.09% on the TMON dataset and RF model achieved 72.88% on the TMO dataset. This work aims to reduce the experimental time for the design and development of an electrocatalyst using a data-driven paradigm.