Clustering stationary and non-stationary time series based on autocorrelation distance of hierarchical and k-means algorithms

Mohammad Alfan Alfian Riyadi (Departement of Statistics, Institut Teknologi Sepuluh Nopember)
Dian Sukma Pratiwi (Departement of Actuarial Science, Bandung)
Aldho Riski Irawan (Departement of Statistics, Institut Teknologi Sepuluh Nopember)
Kartika Fithriasari (Departement of Statistics, Institut Teknologi Sepuluh Nopember)

International Journal of Advances in Intelligent Informatics
Vol 3, No 3 (2017): November 2017

Observing large dimension time series could be time-consuming. One identification and classification approach is a time series clustering. This study aimed to compare the accuracy of two algorithms, hierarchical cluster and K-Means cluster, using ACF’s distance for clustering stationary and non-stationary time series data. This research uses both simulation and real datasets. The simulation generates 7 stationary data models and another 7 of non-stationary data models. On the other hands, the real dataset is the daily temperature data in 34 cities in Indonesia. As a result, K-Means algorithm has the highest accuracy for both data models.