lusccc / dynamic_vehicle_network_clustering

Under different network representation learning methods and clustering methods, we cluster the vehicles in the dynamic vehicle network, and compare the clustering performance.

dynamic_vehicle_network_clustering

Under different network representation learning (NRL) methods and clustering methods, we cluster the vehicles in the dynamic vehicle network, and compare the clustering performance.

Paper: Vehicle Trajectory Clustering Based on Dynamic Representation Learning of Internet of Vehicles (https://ieeexplore.ieee.org/document/9115819)

NRL Methods

• DynWalks
• DeepWalk
• LINE
• node2vec
• GraRep

DynWalks: https://github.com/houchengbin/GloDyNE.

We use the code of the other three NRL methods in OpenNE to learn the embedding vectors of vehicles in the dynamic vehicle network. OpenNE: https://github.com/thunlp/OpenNE

Clustering methods

• KMeans
• HierarchicalClustering
• SpectralClustering
• GaussianMixture
• KMedoids

Metrics

• Silhouette Score
• Dunn Validity Index
• Davies Bouldin Score
• Calinski Harabaz Score

The commands for training the embedding vectors of vehicles using different NRL methods are listed in cmd_command.txt