Abstract | ||
---|---|---|
We understand a long-term time series by transitions of trends and features. We must partition time series into several terms of intervals of different trends and features. In statistics, change detection expresses it as a degree of change using probability distributions. In this paper, we propose a method to partition time series using a hierarchical clustering. First, we have clusters of line segments connecting adjacent data. Then, we unify two adjacent similar clusters into one with a total similarity calculated by the weighted average of three similarities of value, change of values and oscillations. Since the fixed weights cause the partitions that do not fit our sense, we propose dynamical weights with three similarities and sizes of adjacent clusters. Furthermore, in order to exclude small clusters of outliers, we define similarities of two clusters skipping the small cluster. We apply this method to actual time series and show results. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1109/SCIS-ISIS.2018.00139 | 2018 Joint 10th International Conference on Soft Computing and Intelligent Systems (SCIS) and 19th International Symposium on Advanced Intelligent Systems (ISIS) |
Keywords | Field | DocType |
time series,hierarchical clustering,partitioning of time series data | Hierarchical clustering,Computer science,Theoretical computer science,Artificial intelligence,Partition (number theory),Machine learning | Conference |
ISSN | ISBN | Citations |
2377-6870 | 978-1-5386-2634-4 | 0 |
PageRank | References | Authors |
0.34 | 3 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Katsutoshi Takahashi | 1 | 179 | 14.14 |
Motohide Umano | 2 | 183 | 28.91 |
Noriyuki Fujimoto | 3 | 280 | 25.23 |