Ensemble Learning Approach with LASSO for Predicting Catalytic Reaction Rates

Abstract

The prediction of the initial reaction rate in the tungsten-catalyzed epoxidation of alkenes is demonstrated using a machine learning approach. The ensemble learning framework used in this study consists of random sampling with replacement from the training dataset, the construction of several predictive models (weak learners), and the combination of their outputs. This approach enables us to obtain a reasonable prediction model that avoids the problem of overfitting, even when analyzing a small dataset.

Sodium Tungstate Dihydrate (Na2WO4·2H2O): A Mild Oxidizing and Efficient Reagent in Organic Synthesis

“…13 (I) Yada et al described a tungsten-catalysed epoxidation of alkenes. 14 This study allowed the catalytic conditons to be optimized.…”

Incorporating Domain Knowledge and Structure-Based Descriptors for Machine Learning: A Case Study of Pd-Catalyzed Sonogashira Reactions

“…In another study, Doyle and co-workers [ 20 ] used computed atomic and empirical molecular properties (e.g., electrostatic charge, NMR shift) as well as binary categorical identifiers to predict reaction yields for the deoxyfluorination of a broad range of alcohols with sulfonyl fluorides using random forest based on 640 reactions. Yada, Sato, and co-workers [ 21 ] predicted the reaction yields for a tungsten-catalyzed epoxidation of alkene with hydrogen peroxide by coupling DFT-calculated descriptors with the least absolute shrinkage and selection operator methods, a regression analysis method commonly used in machine learning, with 3800 reactions. Denmark and co-workers [ 22 ] used feed-forward neural networks coupled with electronic descriptors and novel 3D steric descriptors generated from DFT calculations to predict the selectivity of phosphoric acid-catalyzed thiol addition reactions using around 1000 data points.…”