Answers are grounded in this paper and cite their sources.
Feature Selection (FS) is a pivotal technique in machine learning (ML) that improves predictive performance, model interpretability, and computational efficiency by reducing data dimensionality and isolating the most informative variables. In the dynamic environment of peer-to-peer (P2P) lending, FS is crucial for accurate credit risk assessment, borrower profiling, and loan default prediction. P2P platforms generate vast and heterogeneous datasets encompassing demographic, financial, behavioural, and transactional information, where redundant or irrelevant features can degrade model accuracy and scalability. This review provides a comprehensive examination of FS methodologies, including filter, wrapper, and embedded approaches, analysing their trade-offs in accuracy potential, computational cost, and interpretability. The study further explores classification models, supervised learning algorithms designed to predict borrower repayment behaviour, covering linear, non-linear, and tree-based ensembles widely applied in credit scoring. Classification methods address critical challenges in P2P lending, such as class imbalance, explainability, and the need for scalable, high-performing predictive systems. By synthesizing recent advances and practical applications, this review offers a structured guide for researchers and practitioners to select FS techniques and classification models aligned with P2P lending’s requirements. Emphasis is placed on optimizing predictive accuracy, enhancing interpretability, and supporting data-driven decision-making to strengthen credit evaluation processes, mitigate default risk, and promote sustainable growth across P2P lending platforms.
No takes yet. Share an insight, caveat, or question.
Atef et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: