BUILD A PERSONALIZED TRAVEL RECOMMENDATION SYSTEM MODULE: COLLABORATIVE FILTERING

Abstract

Recommender is the essential companion for the modern traveler. The Bitap and collaborative travel recommender describe the architecture of a personalized mobile travel predictor by performing content-based filtering and inductive learning techniques depending on travel footprints. Further, the intelligent recommender is powered by collaborative techniques to resolve the multiarmed contextual bandit problem of the exploration/exploitation dilemma a traveler may face in his endeavor. The prediction process helps to estimate the forthcoming user action based on his or her past behaviors, which are considered the best indicators for any future actions. The recommendation process suggests various travel options to the user in line with the inferred behavior. Inductive learning techniques are critical in content-based filtering-based prediction. This model adapts collaborative filtering-based approaches to perform the actual recommendations. In addition to the results obtained via exploitation, an exploration mechanism is also proposed in combination with exploitation so that the best recommendations are presented to the traveler. In this study, a travel recommender system based on predictions of user behavior identified via inductive learning of past transactions is proposed and designed. This system can provide personalized recommendations for users on the move. In the proposed architecture, a location dictionary is maintained with a large pool of data about possible target locations along with their attributes and features. Each visit by the user to any of these locations is recorded and serves as a repository that can be mined to identify the person’s preferences. As soon as the person reaches a new location, recommendations can be provided to him or her based on the past behavior pattern.