A New Approach for Road Type Classification Using Multi-stage Graph Embedding Method

Classifying road types using machine learning models is an important component of road network intelligent systems, as outputs from these models can provide useful traffic information to road users. This paper presents a new method for road-type classification tasks using a multi-stage graph embedding method. The first stage of the proposed method embeds high-dimensional road segment feature vectors to a smaller compact feature space using Deep AutoEncoder. The second stage uses Graph Convolution Neural Networks to obtain an embedded vector for each road segment by aggregating information from neighbouring road segments. The proposed method outperforms the state-of-the-art Graph Convolution Neural Networks embedding method for solving a similar task based on the same dataset.