A Hybrid Recommendation System For Hotel Room Reservation

Hybrid recommendation systems is a combination of different approaches to provide better recommendation system. It's common for anyone to decide the destination based on the trip's intent. To consider trip's intent as a feature of the hotel, the system analyzes text reviews from each hotel and explore it as topic-based trip's intent classification. The system applied Latent Dirichlet Allocation (LDA) to generate the topic probabilities of the corpus. The topics are classified based on the trip's intent consist of romance, business, and adventure. The proposed applied hybrid system is to combine the topic probability model and item-based collaborative filtering. To calculate item-based similarity, the system implemented cosine-based similarity to compute the similarity between two items and treats each item as a vector. The features are composed of the index selected preferences, LDA topic probability value, and number of reviews. In order to improve the accuracy of rating score, this system also analyses the overall rating score using Normal Distribution based Reputation (NDR). The most preferable hotel can be derived from top list of the recommendation list.