Clustering algorithms for off-road vehicles’ usage patterns analysis from CAN bus data

Nowadays, we live in a fast changing society. Many new technologies are created and applied for making our life much more convenient and comfortable. IoT (Internet of Things) is becoming the key technology which can be used for monitoring and improving the producing process and working process. A massive amount of data is collected every day through IoT devices, in this thesis, I apply clustering algorithm on the time series data of construction vehicles for identifying different usage patterns. Two different clustering approaches were applied in this thesis: clustering by distribution and clustering by record. After comparing their performance, clustering algorithms (K-Means, DBSCAN) were performed on several subsets. The results show that K-Means can distinguish different phase and working status of vehicles. The second clustering was performed for gaining more insights of the previous results. After the second clustering, the algorithm can distinguish different workload of vehicles.