Image Driven Optimal Personalized Route Recommendation

Optimal path recommendation relies on a trade-off between the route distance and the places of interest (POIs) present in the route, taking account their number and their value. This abstract problem has multiple subjective solutions. Here, a novel framework for personalized route recommendation is presented aiming at maximizing tourist satisfaction. The optimal path algorithm is driven by POI images which are embedded in a hypergraph modeling the preferences of users for certain POIs. A ranking vector representing the relevance between the POIs and a user is computed and used as an element of the cost vector in the shortest path problem. An encoded graph is employed in order to keep computational complexity at an acceptable level. Moreover, a modified version of the New Approach of Multiobjective A* (NAMOA) search algorithm is proposed. This modified version, coined as Multi Objective NEgative Cyclic A* (MONECA), is a practical solution for multi objective shortest simple path problem in graphs with limited number of nodes, negative weights, and negative cycles.