An XGBoost Algorithm Based on Molecular Structure and Molecular Specificity Parameters for Predicting Gas Adsorption

Li, Lujun; Zhao, Yuejun; Yu, Haibin; Wang, Zhuo; Zhao, Yongjia; Jiang, Mingqi

doi:10.1021/acs.langmuir.3c00255

Cited by 9 publications

(4 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the process of solving the minimum value of the objective function, the constant term has no effect on the result. Therefore, the simplified objective function after removing the constant term is as shown in Equation ( 10) [34]:…”

Section: Methodsmentioning

confidence: 99%

Recognition of Longitudinal Cracks on Slab Surfaces Based on Particle Swarm Optimization and eXtreme Gradient Boosting Model

Liu,

Jiang,

Shi

et al. 2024

Processes

View full text Add to dashboard Cite

Longitudinal cracks are a common defect on the surface of continuous casting slabs, and cause increases in additional processing costs or long-time interruptions. The accurate identification of surface longitudinal cracks is helpful to ensure the casting process is adjusted in time, which significantly improves the quality of slabs. In this research, the typical temperature characteristics of thermocouples at the position of longitudinal cracks and their adjacent locations were extracted. The principal component analysis (PCA) method was used to reduce the dimensions of these characteristics to remove the redundant information. The particle swarm optimization (PSO) method was introduced to optimize the parameter. On this basis, a recognition model of surface longitudinal cracks was established, based on a particle swarm optimization–eXtreme gradient boosting (XGBOOST) model. Finally, this model was trained and tested using longitudinal crack and non-longitudinal crack samples and compared with the decision tree, the gradient boosting decision tree (GBDT) and XGBOOST models. The test results showed that PSO-XGBOOST had the best identification performance in all evaluation indexes. The accuracy, F1 score and alarm rate results were 95.8%, 92.3% and 100%, respectively, and the false alarm rate was as low as 5.5%. The research results provide a theoretical basis and a reliable model for surface longitudinal crack identification.

show abstract

Section: Methodsmentioning

confidence: 99%

Recognition of Longitudinal Cracks on Slab Surfaces Based on Particle Swarm Optimization and eXtreme Gradient Boosting Model

Liu,

Jiang,

Shi

et al. 2024

Processes

View full text Add to dashboard Cite

show abstract

“…Extreme gradient boosting (XGB) is selected as the supervised ML algorithm for this study. XGB is a powerful gradient-boosting implementation that can handle complex, high-dimensional data sets efficiently; its ability to handle nonlinear relationships between features and target variables was particularly advantageous for predicting gas uptake in the MOFs. − The hyperparameters of our XGB models are provided in Supporting Information. We employed the open-source Python toolkits within Scikit-learn for training and validating our ML models.…”

Section: Computational Methodologiesmentioning

confidence: 99%

Engineering Machine Learning Features to Predict Adsorption of Carbon Dioxide and Nitrogen in Metal–Organic Frameworks

Deng,

Sarkisov

2024

J. Phys. Chem. C

View full text Add to dashboard Cite

In this article, we propose simple, interpretable machine learning (ML) features aimed at improving the accuracy of ML models in predicting adsorption of gases, such as methane, ethane, propane, krypton, and xenon, in metal−organic frameworks (MOFs). In particular, we introduce energy-based features that incorporate surface energy histograms and radial distribution functions. These features effectively capture spatial variations in interaction energy, thereby improving the predictive capabilities of the ML models. We further extend this approach to predict the adsorption of carbon dioxide and nitrogen in a diverse set of MOFs under ambient conditions by including the Coulombic energy component in the proposed energy-based features. The final ML model combines the energy-based features with geometric characteristics of MOFs and Henry's law constants. This integration results in significantly increased predictive accuracy of the models, especially within the moderate pressure regime, when compared to that of previous studies. Using the CoRE MOF database for training and testing, our ML model achieves a prediction accuracy of R 2 > 0.96 for methane, ethane, krypton, and xenon at various conditions, R 2 > 0.96 for propane, R 2 > 0.80 for carbon dioxide, and R 2 > 0.97 for nitrogen. Using the CRAFTED database of simulated adsorption isotherms for carbon dioxide and nitrogen, we demonstrate the robustness of the ML model in predicting single-component isotherms at a specific temperature. In this test, the model achieves R 2 > 0.87 for carbon dioxide and R 2 > 0.90 for nitrogen. The physics-based nature of the proposed features allows us to provide some interpretation of why the predictions of the ML model are better for some materials than for others. According to this interpretation, an accurate representation of the shape of the potential energy surface plays an important role in capturing the process of adsorption. Finally, we acknowledge certain limitations of the current model such as the oversimplification in representing the geometry and interactions of molecules with polar interactions.

show abstract

“…Based on the GCMC simulation data set, Lu et al proposed a deep learning classification model powered by the CGCNN for the discovery of the optimal hydrogen-storage MOFs and verified the transferability of the model on the hMOF database. Li et al developed an improved extreme gradient boosting (XGBoost) algorithm based on the graph isomorphic network (GIN) for predicting the adsorption performance of MOFs.…”

Section: Introductionmentioning

confidence: 99%

Graph Transformer with Convolution Parallel Networks for Predicting Single and Binary Component Adsorption Performance of Metal–Organic Frameworks

Zhao,

Gong

et al. 2023

ACS Appl. Mater. Interfaces

Self Cite

View full text Add to dashboard Cite

Metal−organic frameworks (MOFs) are considered one of the most important materials for carbon capture and storage (CCS) due to the advantages of porosity, multifunction, diverse structure, and controllable chemical composition. With the continuous development of artificial intelligence (AI) technology, more and more machine learning models are used to identify MOFs with high performance within a massive search space. However, current works have yet to form a model that uses graphstructured data only, which can predict the adsorption properties of single and binary components. In this work, we proposed and developed a graph transformer, combined with convolution parallel networks, called GC-Trans. The model can accurately and efficiently predict the adsorption performance of MOFs under the single-and binary-component adsorption conditions using only the features of the crystal diagram as inputs. By extracting and fusing local and global feature information, the model has stronger expression and generalization abilities. Thus, we used it to screen the ARC-MOF database and analyze the MOF structures that meet the target requirements. Additionally, to demonstrate the transferability of the model, we applied transfer learning methods to predict the CO 2 /CH 4 separations and CH 4 uptake, both of which showed good predictive performance.

show abstract

An XGBoost Algorithm Based on Molecular Structure and Molecular Specificity Parameters for Predicting Gas Adsorption

Cited by 9 publications

References 48 publications

Recognition of Longitudinal Cracks on Slab Surfaces Based on Particle Swarm Optimization and eXtreme Gradient Boosting Model

Recognition of Longitudinal Cracks on Slab Surfaces Based on Particle Swarm Optimization and eXtreme Gradient Boosting Model

Engineering Machine Learning Features to Predict Adsorption of Carbon Dioxide and Nitrogen in Metal–Organic Frameworks

Graph Transformer with Convolution Parallel Networks for Predicting Single and Binary Component Adsorption Performance of Metal–Organic Frameworks

Contact Info

Product

Resources

About