Personalized Route Recommendation With Neural Network Enhanced Search Algorithm

In this work, we study an important task in location-based services, namely Personalized Route Recommendation (PRR) . Given a road network, the PRR aims to generate user-specific route suggestions for replying users’ queries. A classic approach is adapt search algorithms construct pathfinding-like solutions. These methods typically focus on reducing space with suitable heuristic strategies. For these algorithms, strategies are often handcrafted, which not flexible work complicated settings. addition, it difficult utilize useful context information procedure. To develop more principled solution task, propose improve neural networks solving based widely used $A^{*}$ algorithm. The main idea of our automatically learn cost functions key algorithms. Our model consists two components. First, employ attention-based Recurrent Neural Networks (RNN) from source candidate location by incorporating information. Instead learning single value, RNN component able time-varying vectorized representation moving state user. Second, use estimation network predicting destination. capturing structural characteristics, built top position-aware graph attention networks. components integrated way deriving accurate Extensive experiment results three real-world datasets have shown effectiveness and robustness proposed model.