The real-time time-series data is often unlabelled and the process time is also limited. Therefore, fair assumption of the solution boundaries for that particular data streams is difficult. In this paper, the different clustering approaches are applied to a domain specific application to analyse the variations in the solution outcomes. For this study, an electrical engineering problem of identifying generator coherency in real-time based upon the measurement data is considered. For time-critical applications, a non-ambiguous and effective response is desirable from the coherency identification methods in real-time. In this paper, the four clustering methods namely, hierarchical, k-means, spectral, and fuzzy c-means clustering are applied to the real-time data clustering problem with undefined labels. Performance and suitability of these clustering methods are studied on IEEE 68-bus system with the help of two cluster validity measures.
A Case Study on Clustering Application to Unlabelled Real-time Time-series Data Streams
2021-12-02
2522104 byte
Conference paper
Electronic Resource
English
Learning with constrained and unlabelled data
IEEE | 2005
|Real Output Convergence and Trade Openness: Fuzzy Clustering and Time Series Evidence
Online Contents | 2008
|New Unlabelled Process Oils for the Tyre and Rubber Industry
British Library Conference Proceedings | 1999
|METOCEAN DATA CLUSTERING FOR REAL-TIME MANEUVERING SIMULATIONS
TIBKAT | 2018
|