Comparative Study of Machine Learning Algorithms for Crystal Structure Predictive Modeling of Li-ion Battery Cathodes

Abstract

Identification of the crystal structure of the cathodes of Li-ion batteries is important to improve energy storage technologies. This research has three goals: (1) to determine the important parameters to predict the crystal systems (monoclinic, orthorhombic, and triclinic); (2) to evaluate different machine learning algorithms to make the correct prediction; and (3) to give insights on how to optimize materials used in batteries. We have studied a Kaggle dataset comprising a number of physicochemical descriptors such as chemical formula, energy above the convex hull, band gap, number of atomic sites in the crystal, density, and volume. We evaluate and compare systematically the predictive performance of 10 machine learning algorithms, including Random Forest, Support Vector Machine, Logistic Regression, K-Nearest Neighbor, Decision Tree, Gradient Boosting, XGBoost, Naive Bayes, AdaBoost, and CatBoost. The final model with the best predictive performance resulted from the optimized stacking model that achieved an accuracy of 72.06% and an Receiver Operating Characteristic–Area Under the Curve (ROC–AUC) of 0.8438 after removing leaky features and applying advanced validation methods. These results are consistent with past research that reveals that ensemble learning methods can be used to extract intricate trends in materials information. Nevertheless, this study is based on one dataset with a rather small sample size, which can impact the broader applicability of the findings.   Received: 25 November 2025 | Revised: 17 April 2026 | Accepted: 24 June 2026   Conflicts of Interest The authors declare that they have no conflicts of interest to this work.   Data Availability Statement The data that support the findings of this study are openly available in Kaggle at https://www.kaggle.com/datasets/divyansh22/crystal-system-properties-for-liion-batteries/data. All source code, model implementation, and experimental workfloware available from the GitHub repository at https://github.com/alokchauhan-collab/Crystal-Structure-Predictive-Modeling-of-Li-ion-Battery-Cathodes.   Author Contribution Statement Mega Novita: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization, Project administration. Alok Singh Chauhan: Conceptualization, Methodology, Software, Formal analysis, Writing – review & editing, Visualization, Supervision. Rizky Muliani Dwi Ujianti: Software, Investigation, Data curation, Visualization. Sigit Ristanto: Methodology, Validation, Writing – review & editing, Project administration. Bambang Agus Herlambang: Writing – review & editing. Kazuyoshi Ogasawara: Resources, Writing – review & editing. Chong Geng Ma: Resources. Mikhail Brik: Resources, Writing – review & editing, Supervision. Michal Piasecki: Resources, Writing – review & editing.