Financial Loan Default Risk Prediction Based on Big Data Analysis

In the financial sector, the issue of loan default is critical, and concurrently, the financial lending industry continually contends with market and technological shifts. This study aims to utilize machine learning methods to analyze a substantial amount of bank loan default data. The goal is to develop a loan default prediction model based on the BP neural network, identifying potential default risks at an early stage for proactive risk mitigation. Through coefficient analysis and significance tests, this paper ultimately selects 10 variables as input features for the model. Employing a step-by-step addition approach, the model's performance is evaluated, leading to the determination of a 2-hidden-layer BP neural network. Multiple K-value models are assessed, concluding with a K-value of 4 for the KNN algorithm. Finally, this paper validates the two models—BP neural network and KNN algorithm—analyzing accuracy, recall, and precision. The BP neural network model achieves an accuracy of 70% or more, a recall rate exceeding 55%, and a precision rate surpassing 60%.