A Systematic Review of Machine Learning Algorithms for Classification: General Approaches and Environmental Applications

The field of Machine Learning has seen rapid advancement from 2022 to 2025 due to more cutting-edge computational tools, hybrid models and improved specified techniques. This review has been written to assess the widely used classification algorithms, such as traditional, ensemble-based and deep learning. It evaluates their performance in practical applications. The search covered five open learning databases, which are: Google Scholar, Semantic Scholar, arXiv, DOAJ and ResearchGate. 57 studies published between 2022 and 2025 met the selection criteria. Findings show that Random Forests and XGBoost are effective for structured datasets, and CNNs and transformers are more suitable for unstructured datasets. Hybrid deep learning ensembles are more stable as they can capture spatial and temporal patterns. This review provides a summary of the results, including a comparison table and an outline of areas that require further work.