Personal Credit Evaluation Model Based on Federated Learning

As the economy grows rapidly and consumer attitudes towards spending change, there has been a surge in demand for credit services, prompting the government to relax lending policies. However, this has also led to an increase in personal credit risk, causing the rate of non-performing loans to rise year by year. To address this challenge, experts and scholars have utilized artificial intelligence technology to design a variety of personal credit assessment models, effectively reducing credit risk. Nevertheless, the privacy and sensitivity of financial data have limited data sharing, leading to the phenomenon of data islands, with some small and micro banks potentially lacking sufficient data to train models. This paper introduces a new personal credit assessment model based on federated learning and the DeepFM model, aimed at solving the data sharing issue. It enhances model quality and accuracy through training with federated learning while ensuring data privacy and security.