<p>High-quality personalized travel route planning is a path planning problem that requires comprehensive consideration of path information and user interests. Traditional methods are difficult to plan paths that comprehensively satisfy user interests under multiple constraints with weak generalization ability. Meanwhile, it is still difficult to determine the weight of path information. Aiming at this problem, we propose an optimal path prediction method for keyword-aware based on graph convolutional neural networks to solve the problem of high-quality personalized travel route planning. In this method, a new fuzzy nearest neighbor matching method is proposed to accelerate the mapping of user interests to road network maps for efficient planning of high-quality paths. Then, a path weight scoring mechanism is designed, which contains the capabilities of assigning scores to the path information and simulating the user’s choice preferences for subsequent personalization. The mechanism not only provides ideas for the weights to be difficult to determine but also provides inputs for the features of the incoming neural network. In addition, we designed the graph convolutional network (GCN) prediction network to provide a guarantee for path quality. Through a series of path planning experiments, it is shown that compared with the traditional method, the average query time of this method is reduced by 3.312 s, the average score of path quality is increased by 0.471, and the average improvement of recall is 1.6% as well, which proves the superiority of this method.</p>