Adaptive route optimization in tourism using ant colony graph neural networks with deep learning

In a tourism recommendation system, adaptive route optimisation is an essential requirement for a cost-effective navigation system. When planning a tourism destination, contextual features are essential for creating a model that selects the most efficient route. The existing model fails to account for feature context when determining the path and calculating the shortest distance. The famous Dijkstra algorithm is static and cannot adapt to complex real-world factors such as climate, traffic, and road closures. To automate route optimisation, this research employs deep learning to identify patterns from existing datasets and adapt to the dynamic environment by learning new patterns. Adaptive route optimisation is conducted using an ant colony optimization (ACO) graph neural network (GNN) with a weight factor. ACO operates with pheromone updates for path selection, while GNN connects nodes based on pheromone and other weight factors. This model is less complex and requires less computation time to determine the nearest route. The performance of the model is evaluated, and the proposed model establishes the shortest path in 1.92 s with a cost of 2,141, compared to ACO and ACO with reinforcement learning (ACO-RL).