Scholarlink

Credit risk management is an essential activity to avoid financial stress beforehand. Multiple subjective and quantitative factors such as balance and payments are used to predict credit risks. The prevailing advancements in artificial intelligence have led to the advent of machine learning to generate effective predictive models by using user data. Utilization of data from multiple sources at different locations is useful to create better models. However, data sharing is challenging as it the privacy of users’ sensitive data at risk and incurs huge communication costs. Federated learning is a technique in which an algorithm is trained across numerous decentralized edge devices or servers without sharing original local data samples. In this research work, all the data furnishers participate with personal credit data, and the model is trained on the data silos of multiple financial institutions and banks without even seeing the data rather than the traditional method of sending all the data by data furnishers to the Data Bureaus and then doing analysis. Users of the proposed framework will utilize federated learning to generate credit scores without relying on data suppliers to deliver information. FedAvg method with stochastic gradient descent classification performs optimally in the proposed system. Prior knowledge of the credit score is useful to plan and optimally handle the credit risks to avoid financial stress efficiently.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Personal Credit Score Generator Using Federated Learning for Financial Stress Management

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Ravneet Kaur,
  • Seema Wazarkar,
  • Rohit Jain,
  • Ripul Ahuja,
  • Ishit Bajaj,
  • Saloni Bali

摘要

Credit risk management is an essential activity to avoid financial stress beforehand. Multiple subjective and quantitative factors such as balance and payments are used to predict credit risks. The prevailing advancements in artificial intelligence have led to the advent of machine learning to generate effective predictive models by using user data. Utilization of data from multiple sources at different locations is useful to create better models. However, data sharing is challenging as it the privacy of users’ sensitive data at risk and incurs huge communication costs. Federated learning is a technique in which an algorithm is trained across numerous decentralized edge devices or servers without sharing original local data samples. In this research work, all the data furnishers participate with personal credit data, and the model is trained on the data silos of multiple financial institutions and banks without even seeing the data rather than the traditional method of sending all the data by data furnishers to the Data Bureaus and then doing analysis. Users of the proposed framework will utilize federated learning to generate credit scores without relying on data suppliers to deliver information. FedAvg method with stochastic gradient descent classification performs optimally in the proposed system. Prior knowledge of the credit score is useful to plan and optimally handle the credit risks to avoid financial stress efficiently.