Machine Learning Assisted Selection of Catalyst for γ-Valerolactone Hydrogenation from Levulinic Acid

The preparation of γ-valerolactone by levulinic acid (LA) hydrogenation is green, efficient, and economical, but the traditional catalyst selection and optimization methods have low efficiency and high cost, which cannot meet the development needs of the chemical industry. To this end, we conducted new research to develop a machine learning framework to predict LA conversion and γ-valerolactone yield, accelerating catalyst selection and optimization. Through K-means clustering preliminary classification data sets, composite minority oversampling technology and adaptive composite sampling resampling unbalanced data sets were used to solve the problem of small sample size and improve classification performance. The performance of four machine learning optimization algorithms was evaluated, and the superior performance of the support vector machine was found to be the core of the model. We not only pursue prediction accuracy but also find that reaction temperature was the main influencing factor through the shapley additive explanation. The most potential catalyst Ru/N@CNTs was selected based on the feature importance analysis results combined with the genetic algorithm multiobjective optimization model.