High performing metal oxide based materials identified for symmetric pseudocapacitor electrodes using unsupervised machine learning

Supercapacitors proved to be good alternatives to batteries in the electrochemical energy storage because of their high power and extended lifespan. Recently an extensive database on supercapacitor electrode materials and its performance parameters was reported by Soni et al. which was generated using Web Scrapping and Natural Language Processing. The subset of that database is used in this study to find out the lead candidate electrode materials by clustering using unsupervised machine learning techniques. The subset consists of electrode materials made up of metal oxides and their composite electrode materials used in supercapacitors. Two types of clustering have been performed namely K-means and Hierarchical clustering to identify the inherent clusters present in the dataset. It has been noted that the maximum Silhouette Score of 0.4755 and 0.4513 is shown by K-mean and Hierarchical clustering with three and seven clusters respectively. This study reveals the one cluster which contains Co 3 O 4 , CeO 2 , CuMnO 2 , MnO 2 , TiO 2 and RuO 2 based composites as the electrode materials showing exceptionally good Specific Capacitance and Cyclability. This is first of its kind application of unsupervised machine learning techniques in the realm of supercapacitor research to attempt clustering the electrode materials. This attempt identifies patterns in the potential electrode materials and thus reduces the complexity of material selection for a particular performance requirement.