In this research, we present a traffic analysis zones division model based on complex network theory, which is crucial for conducting feasibility studies for new road projects. The model is designed with three fundamental components: a multi-layer network, feature extraction, and a two-step division method. The highway network is represented using the multi-layer approach that includes the travel cost network, road structural network, and Origin-Destination network. Additionally, to measure the similarity between nodes, both structural attributes and clustering attributes are considered as eigenvalues in the similarity measurements. Furthermore, the two-step division method comprises Louvain Community Detection algorithm for coarse division and modified clustering algorithm for fine division. With the proposed model, the nodes of highway can be divided into traffic analysis zones. Through the case study of the Guangdong-Hong Kong-Macao Greater Bay Area, the results of the model demonstrate a significant improvement in three evaluation metrics when compared to other methods, highlighting the superiority of the proposed model.