Investment decision making for large-scale Peer-to-Peer lending data: A Bayesian Neural Network approach

Abstract

Peer-to-Peer (P2P) lending, as a pivotal innovation in the financial sector, presents both significant opportunities and complex challenges for portfolio management. This study introduces an advanced P2P lending portfolio optimization model that integrates Bayesian Neural Networks (BNNs) within the Neural Additive Model (NAM) framework to address these challenges. Our primary objective is to enhance the interpretability and operational efficacy of large-scale P2P lending portfolios. By leveraging BNNs, we not only predict returns but also quantify uncertainty to assess loan risks effectively. To augment the model’s transparency, NAMs are employed to elucidate the impact of various features on investment outcomes. Subsequently, a genetic algorithm optimizes the allocation of investment weights, ensuring maximum profitability. The proposed strategy is validated using real-world P2P lending data, demonstrating superior performance compared to traditional benchmarks in predicting P2P lending profits. Empirical evidence suggests that our approach significantly enhances investment returns by facilitating informed decision-making. This research provides actionable insights for investors in the P2P lending domain and represents a substantial advancement in risk management and decision-making through the innovative application of BNNs and NAMs.