Application of Soft Computing Represented by Regression Machine Learning Model and Artificial Lemming Algorithm in...

Area of Science:

• Materials Science
• Chemical Engineering
• Computational Chemistry

Background:

• Metal-organic frameworks (MOFs) possess unique properties making them promising for hydrogen storage applications.
• Accurate prediction of hydrogen storage capacity in MOFs is crucial for material selection and process optimization.

Purpose of the Study:

• To develop and evaluate several regression-based machine learning models for predicting hydrogen storage capacity in MOFs.
• To optimize machine learning model hyperparameters using the artificial lemming algorithm (ALA).

Main Methods:

• Development of artificial neuron network (ANN), support vector regression (SVR), random forest (RF), extreme learning machine (ELM), kernel extreme learning machine (KELM), and generalized regression neural network (GRNN) models.
• Hyperparameter optimization using the artificial lemming algorithm (ALA).
• Model training and testing using experimental hydrogen storage data, with performance evaluation via statistical metrics, regression plots, and Taylor graphs.

Main Results:

• The ALA-optimized random forest (ALA-RF) model exhibited the highest predictive accuracy.
• Optimal performance metrics for ALA-RF include R² of 0.9845 (training) and 0.9840 (testing), RMSE of 0.2719 (training) and 0.2828 (testing).
• Pressure was identified as the most influential feature for predicting hydrogen storage capacity in MOFs.

Conclusions:

• The ALA-RF model offers a robust and accurate approach for predicting hydrogen storage in MOFs.
• These findings facilitate intelligent selection of MOFs and optimization of hydrogen storage operational conditions.