Work

Machine-learning prediction of freezing temperature in Pb-contaminated soils remediated with biochar

Year: 2026

Type: article

Abstract: Throughout the Pb-contaminated soil rehabilitation procedure, biochar may change the structure of the soil, which has a significant impact on soil freezing. This study examined the change in freezing temperature ( $ToF$ ) during the biochar remediation of Pb-contaminated soils using machine learning techniques, such as Light Gradient Boosting ( $LGB$ ), Extreme Gradient Boosting ( $XGB$ ), Extra Tree Regression ( $ETR$ ) and Random Forest Regression ( $RFR$ ). The data used in this research is derived from previously conducted trials. The system's reliability is significantly impacted by hyperparameters, which must be selected using metaheuristic optimization approaches. To accomplish this, a variety of algorithms are used, including the Catch Fish Optimization ( $CFO$ ). The information, which includes 360 ToF data points, came from studies that were submitted for publication. After studying each element, the study group came to the decision that 80% (288) of the dataset would be utilized for training and model construction, while the other 20% (72) of the dataset would be utilized for validating the model. Randomization, normalization, cross-validation, scaling, and feature significance evaluation were some of the pre-processing methods used to the dataset. The prediction efficacy of $ETR(CFO)$ and $RFR(CFO)$ in predicting ToF was assessed using metrics. With the data, ToF can be reliably predicted by both $ETR(CFO)$ and $RFR(CFO)$ over other baseline models. The assessment process assessed the reliability and plausibility of each model using evaluation criteria, logical analysis, ranking score, and model difference percentage. In achieving its primary target, the $RFR(CFO)$ model performs better than the alternative model.

Source: Scientific Reports

DOI: 10.1038/s41598-026-77024-4

Author: Laila Safi

Institution: Paktia University

Language: English

Topic: Heavy metals in environment

Subfield: Pollution

Field: Environmental Science

Domain: Physical Sciences

SDGs: Life on land, Climate action

Open Access status: gold

Estimated APC paid: $2,690