The Engineering of Bank Recommendation Service Using Machine Learning

Recent trends in banking products development designed to increase the interoperability of systems and the meet of consumers needs. One of the fast growing ways to improve the application of IT services is the deployment of recommender systems - tools, which might automatically generate proposals for goods and services based on the transaction statistics. This is a well-studied research field, which has successful applications in the industry. Recommendation systems have become an integral part of many of today's Internet applications, proving its effectiveness in improving the user experience and sales increase. After the transformation of checks in electronic form, analyzing the banks` and their customers` needs, there is a potential to create really personalized service to customers, involving data from brick-and-mortar retailers and banks. The practical side of such service is the ability to improve the user experience, to enhance knowledge of the bank's clients and create an open service to utilize data with the help of third-party developers and save money from card issuing and acquiring. The paper discusses the algorithms to select the best type of product offer and the most relevant offers based on the methods of collaborative filtering. These algorithms are implemented for a scalable software platform Apache Spark using the Scala and R languages. The experiments aimed to evaluate the quality and scalability of implementation performed on the Microsoft Azure cloud service.