Given the vast amount of information, including the numerous points-of-interest (POIs) and the various hotels, available on travel websites such as tripadvisor or booking, a recommender system would help users, who are planning their next trip, filter out unnecessary information based on their requirements. We improved our previous work on a recommendation system that was intended to facilitate the generation of daily travel itineraries. We used the X-Means clustering algorithm to divide all attraction sites and hotels into groups according to geographical location. Meanwhile, a Word2Vec model was trained using the Wikipedia text corpus to obtain similar tags of specific ones. A tag-based mapping algorithm was applied to create a list of candidate attractions that best match with the user's favorite spots. Finally, by taking into account the weather information, our recommender can further refine the list of candidate attractions and work out a daily itinerary that involves desirable hotels and attractions. The shortest itinerary (SI) and the itinerary with the highest performance/price ratio (MEI) will then be produced for user selection.  The results of a series of experiments demonstrated that, compared to others, our personalized recommender for travel planning can provide a more appealing and detailed travel plan containing daily itineraries for users.