Predicting Methanol Space-Time Yield from CO₂ Hydrogenation Using Machine Learning - Statistical Evaluation of Penalized Regression Techniques

Aug 6, 2026

This study evaluates machine learning penalized regression models—Ridge Regression, Lasso Regression, and Elastic Net Regression—for predicting methanol space-time yield (STY) from CO₂ hydrogenation data. A Cu-based catalyst-derived dataset is preprocessed using cleaning, imputation, outlier removal, and normalization. Model performance is assessed via 10-fold cross-validation and testing on unseen data. Ridge Regression achieves the lowest errors (RMSE 0.7706, MAE 0.5627, MSE 0.5938), while Lasso and Elastic Net show higher metrics, and feature importance highlights GHSV and support molar mass as key factors.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Introduction

Machine learning for CO₂ hydrogenation modeling

Methods

Data preprocessing

Penalized regression models

Model evaluation

Results and Discussion

Error metrics comparison

Feature importance analysis

Conclusion

Practical implications for catalyst optimization

International Journal of Advances in Data and Information Systems Vol. 5, No. 2, October 2024, pp. 216~228 ISSN: 2721-3056, DOI: 10.59395/ijadis.v5i2.1341 r 216 | Predicting Methanol Space-Time Yield from CO₂ Hydrogenation Using Machine Learning: Statistical Evaluation of Penalized Regression Techniques
Harun Al Azies1,2, Muhamad Akrom1,2, Setyo Budi3,2,
Gustina Alfa Trisnapradika1,2, Aprilyani Nur Safitri1,2
1Study Program in Informatics Engineering, Faculty of Computer Science, Universitas Dian Nuswantoro, Indonesia 2Research Center for Quantum Computing and Materials Informatics, Faculty of Computer Science, Universitas Dian
Nuswantoro, Indonesia
3Study Program in Information Systems, Faculty of Computer Science, Universitas Dian Nuswantoro, Indonesia | | | --- | --- | | Article Info
Article history:
Received Sep 09, 2024 Revised Sept 29, 2024 Accepted Oct 30, 2024
Keywords:
Penalized Regression Ridge Regression Methanol Production CO₂ Hydrogenation Lasso Regression Elastic Net Regression
Corresponding Author: | This study investigates the effectiveness of machine learning techniques, specifically penalized regression models Ridge Regression, Lasso Regression, and Elastic Net Regression in predicting methanol space-time yield (STY) from CO₂ hydrogenation data. Using a dataset derived from Cu-based catalyst research, the study implemented a comprehensive preprocessing approach, including data cleaning, imputation, outlier removal, and normalization. The models were rigorously evaluated through 10-fold cross-validation and tested on unseen data. Ridge Regression outperformed the other models, achieving the lowest Root Mean Squared Error (RMSE) of 0.7706, Mean Absolute Error (MAE) of 0.5627, and Mean Squared Error (MSE) of 0.5938. In comparison, Lasso and Elastic Net Regression models exhibited higher error metrics. Feature importance analysis revealed that Gas Hourly Space Velocity (GHSV) and Molar Masses of Support significantly influence catalytic activity. These findings suggest that Ridge Regression is a promising tool for accurately predicting methanol production, providing valuable insights for optimizing catalytic processes and advancing sustainable practices in chemical engineering.
This is an open access article under the CC BY-SA license.

ABSTRACT | | Harun Al Azies,
Study Program in Informatics Engineering, Universitas Dian Nuswantoro,
127 Imam Bonjol Street, Semarang 50131, Indonesia. [Email: harun.alazies@dsn.dinus.ac.id](Email: harun.alazies@dsn.dinus.ac.id) | | 1. INTRODUCTION The industrial production of carbon dioxide (CO₂) as a feedstock to manufacture value-added chemicals like methanol has gained interest in recent years due to its potential to ameliorate climate change and reduce dependency on fossil fuels [1] . CO₂, a major greenhouse gas, contributes heavily to global warming [2], and its conversion into useful chemicals, such as methanol[3], offers a sustainable approach to reducing its impact on the environment [4] . Methanol itself has a broad range of applications, including its use in the production of polymers [5], fuels, and various organic compounds, making it an essential component of modern industrial processes [6],[7] . As the global demand for cleaner energy solutions grows, the need for efficient and sustainable methods of methanol production becomes even more critical [8] . Despite the potential benefits, the CO₂ hydrogenation process to methanol presents significant challenges [9]. The reaction involves a complex interplay offactors, such as catalyst type, temperature, pressure, and reactant ratios, all of which must be optimized to achieve efficient conversion. Copper (Cu)-based catalysts, known for their high activity and selectivity in producing methanol, are widely used for this purpose. However, achieving optimal conditions for maximum methanol yield is complicated due to the nonlinear nature of these interactions [10], [1

FAQs

Which machine learning models are used to predict methanol space-time yield (STY)?

The study uses penalized regression models: Ridge Regression, Lasso Regression, and Elastic Net Regression to predict STY from CO₂ hydrogenation data.

How are the models evaluated in the study?

Models are evaluated using 10-fold cross-validation and then tested on unseen data to measure prediction performance.

What model performed best, and what metrics support this?

Ridge Regression performs best, reaching the lowest RMSE (0.7706), MAE (0.5627), and MSE (0.5938) compared with Lasso and Elastic Net.

Aug 6, 2026

Upload by Emma Wilson