Abstract

Predicting soil toxicity is crucial for assessing environmental risks and safeguarding ecosystems and human well-being. This study conducts an extensive comparative analysis between two robust machine learning algorithms, Random Forest (RF) and Support Vector Machine (SVM), to forecast soil toxicity. Employing a diverse dataset encompassing soil samples from various geographical locations, we examine how effectively RF and SVM models classify soil samples into toxic and non-toxic categories. Our investigation commences with a comprehensive exploration of feature selection methods aimed at identifying the most pertinent predictors for soil toxicity. Subsequently, we train and assess RF and SVM models using these chosen features, employing stringent cross-validation techniques to ensure the reliability and applicability of our findings. Performance metrics such as accuracy, precision, recall, and F1-score are employed to evaluate the predictive capabilities of each model. The outcomes of our study provide intriguing insights into the relative effectiveness of RF and SVM in predicting soil toxicity. While both models exhibit commendable performance, our analysis uncovers subtle differences in their predictive strengths and weaknesses across various soil types and toxicity levels. Furthermore, we delve into the interpretability of model predictions, elucidating the underlying factors influencing soil toxicity and the decision-making process of machine learning models. Ultimately, this research contributes to the advancement of soil toxicity prediction by furnishing valuable empirical evidence on the relative performance of RF and SVM models. The implications of our findings are significant for environmental scientists, policymakers, and stakeholders engaged in soil management and remediation endeavors.

Keywords:Soil toxicity prediction; Machine learning algorithms, Random Forest (RF); Support Vector Machine (SVM); Comparative analysis; Environmental risk assessment; Feature selection and Cross-validation techniques