Advanced hybrid machine learning based modeling for prediction of properties of ionic liquids at different temperatures
Abstract Ionic liquids have attracted much attention in different fields such as chemical and pharmaceutical industries for green processing. Their physicochemical properties should be estimated via advanced methods to save time and costs of experimentation. The main aim of this work is to study thr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | Scientific Reports |
| Subjects: | |
| Online Access: | https://doi.org/10.1038/s41598-025-04450-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Ionic liquids have attracted much attention in different fields such as chemical and pharmaceutical industries for green processing. Their physicochemical properties should be estimated via advanced methods to save time and costs of experimentation. The main aim of this work is to study three significant machine learning models to estimate the surface tension of ionic liquids via advanced hybrid computational techniques. Several advanced models including Decision trees (DT), and two ensembles based on it including extra trees (ET) and random forest (RF) are employed to estimate the properties of ionic liquids at different conditions. The harmony search (HS) algorithm was employed as an optimization algorithm to find the best hyper-parameters combination by minimizing the error. With R2 metric, DT, ET, and RF have values of 0.894, 0.979, and 0.945, respectively. Considering the MAPE error rate, DT, ET, and RF have errors of 4.59E-02, 2.05E-02, and 2.59E-02, respectively. The main model in this study was chosen to be ET because of its less errors and higher R2. |
|---|---|
| ISSN: | 2045-2322 |