A travel recommendation system based on social media activity provides a customized place of interest to accommodate user-specific needs and preferences. In general, the user's inclination towards travel destinations is subject to change over time. In this project, we have analyzed users' twitter data, as well as their friends and followers in a timely fashion to understand recent travel interest. A machine learning classifier identifies tweets relevant to travel. The travel tweets are then used to obtain personalized travel recommendations. Unlike most of the personalized recommendation systems, our proposed model takes into account a user's most recent interest by incorporating time-sensitive recency weight into the model. Our proposed model has outperformed the existing personalized place of interest recommendation model, and the overall accuracy is 75.23%.

Personalization of recommender systems enables customized services to users. Social media is one resource that aids personalization. This study explores the use of twitter data to personalize travel recommendations. A machine learning classification model is used to identify travel related tweets. The travel tweets are then used to personalize recommendations regarding places of interest for the user. Places of interest are categorized as: historical buildings, museums, parks, and restaurants. To better personalize the model, travel tweets of the user's friends and followers are also mined. Volunteer twitter users were asked to provide their twitter handle as well as rank their travel category preferences in a survey. We evaluated our model by comparing the predictions made by our model with the users choices in the survey. The evaluations show 68% prediction accuracy. The accuracy can be improved with a better travel-tweet training dataset as well as a better travel category identification technique using machine learning. The travel categories can be increased to include items like sports venues, musical events, entertainment, etc. and thereby fine-tune the recommendations. The proposed model lists 'n' places of interest from each category in proportion to the travel category score generated by the model.

No abstract

No abstract

No abstract