Rapidly evaluating electrochemical performance of transition metal disulfides for lithium ion batteries using machine learning classifier

Transition metal disulfides (TMS₂) have shown high promise as lithium-ion battery electrodes with capacities ranging from below 200 to over 1000 mAh g⁻¹. However, current machine learning (ML) approaches for performance prediction rely on computationally intensive density functional theory descriptors that are rarely reported in experimental literature. This creates a gap between theoretical predictions and practical screening needs. In this work, we developed a ML classifier using experimental literature data from 218 TMS₂ electrode samples. Four universally available features were extracted: transition metal electronegativity, morphology category, voltage window, and current density. The Gradient Boosting Classifier achieved optimal performance with Area Under the Receiver Operating Characteristic Curve of 0.97 for predicting five capacity grades. Experimental validation using synthesized flower-like MoS₂ and MoS₂, CoS₂, and VS₂ nanoparticles without defined morphology confirmed predictions across multiple current densities. This classification approach provides a practical screening tool that overcomes data availability limitations in literature-based materials discovery.