Travel Package Recommendation Based on Reinforcement Learning and Trip Guaranteed Prediction

Trip planning research and travel package recommendation benefit from current trends in Location Based Social Networks and trajectory related sites nowadays. Travel package recommendation requires the extraction of characteristics of points of interest and setting up a ranking method. Traditional research used to rely on questionnaires without statistical validation methodologies. We proposed a recommendation framework based on reinforcement learning. To reach the objective of generating successful travel packages, we introduced a reward function for ranking points of interest. Based on labeled travel package data provided by travel agencies, two trip guaranteed prediction methods (deep learning and trajectory similarity) were used for travel guarantee prediction. The results of the accuracy and performances of these methodologies showed the prediction models are reliable. We found no statistically significant difference between the recommended and the uncancelled package groups.