Designing a Model for Improving Banking Recommender Systems Based on Predicting Customer's Interests: Application of Data Mining Techniques

Nowadays, banks require new devices such as recommender systems to attract and preserve customers. Unlike most recommender systems in which the given recommendation is based on similarities between the preferences of users, this research has employed the classification techniques where customer’s past interests is considered as the most important feature to provide proper banking services for them. In this research, four classifiers including MLP, SVM, KNN, and Naïve Bayes have been used. Firstly, the data set which was related to the services used by different bank customers was pre-processed and four different classification methods were trained by using it. Then, their validations were assessed by the 10-fold cross validation and the best method was selected. Lastly, the final recommender system which was a combination of four classification methods including Naïve Bayes with performance P=%85.4, 5-nn with P=%83.3, MLP with P=%81.4, and MLP with P=%92.6 respectively proposed for recommendation of four banking services including the internet, mobile, internet transfer and paying on the phone is.