Retail banking is the bedrock of most banks in Kenya. The competition landscape in the sector has evolved considerably over the years, with profound implications for efficiency and stability for the banking sector. Innovations continue to alter the landscape in the banking sector and are envisioned to shape the evolution of credit allocation and delivery of services. With financial institutions having customer bases of thousands and even millions of customer and multiple products, recommending products to customers, cross selling and upselling has been a challenge. The paper developed a recommender system to point a user towards useful and not-yet-experienced retail banking products. The approaches in developing recommendations systems were reviewed. After weighing the limitations in each approach in terms of data required and data protection issues collaborative filtering using implicit feedback from transaction data has been identified, implemented and assessed for effectiveness. The value of a recommender system based on machine learning is that there is no need for expensive data aggregation, recommendations are real-time and data sharing is eliminated hence customer data protection. In conclusion, the model can be optimized by inclusion of more detailed transaction data within the legal framework of customer data protection.
Keywords: Recommender systems, collaborative filtering, machine learning