Title:
Personalized Savings Product Recommendations Using a Hybrid Approach

Authors:
Bayarmaa Dashnyam, Gerelt-Od Uvgunkhuu, Oyundelger Batkmunkh, Mongolia

Abstract:

Financial sources are crucial for the stable business success of financial institutions, and banks actively offer savings products to customers to enhance their sources. Therefore, attracting customers by providing the right products that align with their preferences is essential. A recommendation system helps customers select suitable products and has demonstrated business value through increased sales and institutional income. While recommendation systems have found success in areas like e-commerce and entertainment, implementing them in the financial sector remains challenging due to a lack of explicit customer feedback such as ratings and reviews. This study addresses this gap by developing a hybrid recommendation system specifically for a bank’s savings products. By utilizing transaction-based implicit feedback and combining content-based filtering with collaborative filtering, we have designed a model tailored for the banking context.  We used real banking data from Mongolia covering the period 2021–2023. Among individual models, the content-based method showed the highest accuracy. The hybrid model outperformed both content-based and collaborative filtering approaches when the content weight α = 0.85, achieving the lowest RMSE of 1.260 and MAE of 0.980. This result demonstrates the effectiveness of a well-balanced hybrid approach in predicting customer preferences in banking savings products.  This research contributes to bridging the gap between recommendation technologies and the financial domain, providing practical implications for personalized financial product offerings where traditional feedback is absent.

DOI: https://doi.org/10.51505/IJEBMR.2025.9525