NTRS: A NEW TRAVEL RECOMMENDATION SYSTEM FRAMEWORK BY HYBRID DATA MINING

This study suggests a new travel recommendation system (NTRS) that was developed to generate alternative travel destinations for customers. The proposed approach employs hybrid data mining methods on NTRS by combining classification and clustering algorithms. NTRS can be used for different travel data resources to find the best prediction model to generate accurate recommendations. NTRS was tested by using real travel data which contains flight and hotel bookings. Before applying data mining algorithms, data set was cleansed, grouped and preprocessed. Then classification techniques; ANFIS, RBFN and Naïve Bayes were combined with X-means and Fuzzy Cmeans clustering algorithms to find the best prediction model for proposing alternative trips via NTRS. To identify the most suitable prediction model; recall, specificity, precision, correctness, and RMSE scores were benchmarked and the best one was dynamically selected. According to the testing scenario results, ANFIS and X-means combination scored the finest RMSE and correctness values. Based on the proposed approach's algorithm, travel locations including trip durations and airline companies were generated as recommendation output of the testing scenario. Generated recommendation items can be used for providing suggestions for individuals or it can be used by travel agencies for planning travel campaigns for target traveler groups. NTRS proves that it can be executed for different data sets with hybrid data mining methods.