Prediction of COF Synthesis Reaction Conditions Based on Data Mining and Machine Learning
Lists
Only you see this
Abstract
Covalent organic frameworks (COFs) have demonstrated superior physical/chemical properties and application potential, while the use of machine learning (ML) methods to predict COF synthesis parameters has not been fully explored. This work developed an ML prediction network model for predicting COF synthesis conditions using monomers as input, which guided the synthesis of real COF materials and their application in NIR-II photothermal antibacterial applications. First, we accurately extracted COF synthesis conditions from 609 references using the large language model Deepseek-R1 and established a COF synthesis database SYN-COF. Subsequently, the XGBoost-based ML model was trained and optimized using the COF synthesis database, achieving R2 values of 0.88 for the synthesis temperature and 0.47 for the synthesis time, and demonstrated higher accuracy in competition with 7 human experts. With the assistance of the ML prediction network model, we successfully synthesized previously unreported COF materials, leading to a good NIR-II photothermal antibacterial performance after iodine doping.
BibTeX
@inproceedings{zhao2025prediction,
author = {Jinjin Zhao and Mengwen Yan and Zeyu Shi and Chao Zhao and Lei Qi and Qing Hao and Hong Liu},
title = {{Prediction of COF Synthesis Reaction Conditions Based on Data Mining and Machine Learning}},
booktitle = {ACS},
year = {2025},
doi = {10.1021/acsapm.5c03780},
}Loading related content...