Đề án Thạc sĩ: Xây dựng hệ khuyến nghị về sản phẩm vay cho khách hàng ở công ty tài chính sử dụng deep reinforcement learning

The increasing volume of data and diversity of user preferences have made effective information recommendation a critical challenge, particularly for Recommender Systems (RS) aiming to optimize user experience and engagement. Ensuring accuracy and efficiency in these systems is paramount. Concurrently, Reinforcement Learning (RL) and its advanced variant, Deep Reinforcement Learning (DRL), offer powerful paradigms for modeling human-like learning from experience, demonstrating flexibility and robust learning capabilities that can significantly enhance RS performance. In the current competitive economic climate, recommending suitable loan products is vital for financial companies to meet diverse customer needs and foster financial growth. This research addresses these critical areas by proposing a DRL-based framework to build a personalized loan product recommendation system.

Researchers and practitioners in Artificial Intelligence, Machine Learning, and Recommender Systems, especially those focusing on financial applications; data scientists and product managers in financial technology (FinTech) companies; and academics interested in the application of Deep Reinforcement Learning for personalized services.

This academic work presents a comprehensive approach to developing a sophisticated Deep Reinforcement Learning (DRL)-based recommender system for financial loan products, specifically tailored for customers of a financial company. The core objective is to overcome the limitations of traditional recommender systems by leveraging the dynamic and adaptive capabilities of DRL to provide highly personalized and effective loan recommendations. The methodology involves a thorough theoretical review of Reinforcement Learning (RL) and DRL, exploring their foundational principles, strengths, and existing applications, particularly within the context of Recommender Systems (RS). A key aspect of the research is the construction of a specialized dataset comprising customer profiles and various loan product attributes, potentially drawing from real-world data such as that of Tín Việt financial company, to adequately train the DRL models. Furthermore, the study will utilize simulation platforms like RecSim to validate and evaluate the system's performance. The research aims to develop a system capable of accurately predicting customer interest and proactively suggesting optimal loan products, thereby enhancing user satisfaction and engagement. The practical implications are significant for financial institutions, offering a robust tool to optimize their service delivery, improve customer retention, and gain a competitive edge in a saturated market. By integrating advanced artificial intelligence techniques, this system promises to deliver a more nuanced and responsive approach to financial product recommendation, adapting to evolving customer behaviors and market conditions while improving predictive accuracy compared to previous studies.