[1]张茜茜,陈 平.基于图卷积的离子液体 CO2 溶解度可解释性预测[J].计算机技术与发展,2024,34(02):134-141.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 020]
　ZHANG Qian-qian,CHEN Ping.Interpretable Prediction of CO2 Solubility of Ionic Liquids Based on Graph Convolution Neural Network[J].,2024,34(02):134-141.[doi:10. 3969 / j. issn. 1673-629X. 2024. 02. 020]

点击复制

基于图卷积的离子液体 CO₂ 溶解度可解释性预测()

分享到：

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

34

期数:

2024年02期

页码:

134-141

栏目:

人工智能

出版日期:

2024-02-10

文章信息/Info

Title:

Interpretable Prediction of CO₂ Solubility of Ionic Liquids Based on Graph Convolution Neural Network

文章编号:

1673-629X(2024)02-0134-08

作者:

张茜茜¹; 2 ; 陈 平¹; 2

1. 中北大学 信息与通信工程学院,山西 太原 030051;

2. 信息探测与处理山西省重点实验室(中北大学),山西 太原 030051

Author(s):

ZHANG Qian-qian1; 2 ; CHEN Ping1; 2

1. School of Information and Communication Engineering,North University of China,Taiyuan 030051,China;

2. Shanxi Province Key Laboratory of Information Detection and Processing,North University of China, Taiyuan 030051,China

关键词:

图卷积神经网络; 离子液体; 性质预测; 溶解度; 可解释性

Keywords:

graph convolutional neural networks; ionic liquids; property prediction; solubility; interpretability

分类号:

TP181

DOI:

10. 3969 / j. issn. 1673-629X. 2024. 02. 020

摘要:

为构建离子液体的 CO₂ 溶解度的准确预测模型,考虑到传统模型存在的描述符计算复杂、成本高、关联结构与性质困难、结构特征提取不充分等问题,提出一种融合了加入注意力机制的图卷积神经网络和 XGBoost 的预测模型( APGCN-XGBoost) 。 对 9 897 组离子液体的 CO₂ 溶解度数据的分析结果显示,所提出的 APGCN-XGBoost 模型在预测性能上优于传统的分子指纹模型和图卷积神经网络模型。 此外,通过注意力池化层与 SHAP 方法对模型进行解释,APGCN-XGBoost 模型学习到了离子液体中各个原子和结构的特征信息与分子非局部信息,这些特征信息不仅可以用于性质预测,还可以用于探索化学结构与性质之间的联系,即通过模型的解释,筛选出对于溶解度预测重要的离子液体结构信息,从而实现 CO₂捕获过程中理想离子液体的计算机辅助设计和筛选。

Abstract:

In order to build an accurate prediction model of CO₂ solubility of ionic liquids, considering the problems existing intraditional models,such as complex descriptor calculation,high cost,difficult related structure and properties,and insufficient structuralfeature extraction, a prediction model APGCN - XGBoost combining graph convolution neural network and?

XGBoost with attentionmechanism was proposed. The analysis results of CO₂ solubility data of 9 897 groups of ionic liquids show that the predictionperformance of the proposed APGCN-XGBoost model is better than that of the traditional molecular fingerprint model and graph convolutional neural network model. In addition,the APGCN-XGBoost model learned the characteristic information and molecular non-localinformation of each atom and structure in ionic liquids,which can be used not only to predict the properties,but also to explore the relationship between the chemical structures and properties,that is,to screen out the structural information of ionic liquids that is important forsolubility prediction through the interpretation of the model,thus realizing the computer-aided design and screening of ideal ionic liquidsin the process of CO₂ capture.

常用功能