Ürün önerisi sistemleri için banka müşterilerinin hareketlilik özellikleri ile ürün kullanımı arasındaki ilişkinin araştırılması

Conventional financial decision support systems are based on distinct individual attributes like gender, marital status, age and occupation, while being unaware of the spending habits or spatio-temporal mobility of the individuals. The rapid growth of mobile payment and geo-aware systems, and the emergence of big data present the opportunity to investigate human consumption patterns across space and time. We analyze a one-year transaction dataset of a leading bank in Turkey to understand the relation between mobility behaviors and product usage patterns for Individual Consumer Loan and Retirement Savings products. After data preprocessing, we filter the customers who provide more information in terms of mobility. In order to evaluate the product recommendation model in different circumstances, we divide the data into different scenarios with respect to the last 3 months' product usage pattern, time period, demographic, financial and mobility features. We conduct computational experiments to determine the best feature selection methods and the resulting features, and the best classification algorithms for each dataset independently. Finally, we analyze the classification results and report on most interesting product usage patterns and customer characteristics.