چكيده به لاتين
Designing customer behavior prediction models using data mining algorithms enables banks to identify and timely solve problems before turning their customers around for a variety of reasons. In this study, two models are proposed to predict the loyalty of the receivers of the terminal terminals based on their transactions. These models are based on the standard CRISP-DM methodology and the required data are extracted from the databases of one of the country's private banks. The Clementine 14 software was used to analyze the data, using knowledge-based discovery techniques such as clustering with k-means algorithm, based on important features such as: number of transactions in the current month, number The transaction was clustered last month to determine the status of each of the terminals in each of the months of the year. Key indicators were identified from existing data and with their help, new indicators called loyalty indicators were defined. Finally, with the help of prediction techniques (stepwise regression and decision tree), the loyalty rate of each of the recipients was determined. This research is descriptive in terms of purpose, descriptive and in terms of data collection, documentary evidence. Designed models enable PSPs to identify, based on predictions, customers who are likely to migrate to other competing banks and take the necessary precautionary measures.
