In this study, three algorithms are presented for time series segmentation. The first algorithm is based on the branch-and-bound approach, the second on the dynamic programming while the third is a modified version of the latter into which the remaining cost concept of the former is introduced. A user-friendly computer program called AUG-Segmenter is developed. Segmentation-by-constant and segmentation-by-linear-regression can be performed by the program. The program is tested on real-world time series of thousands of terms and found useful in performing segmentation satisfactorily and fast.
Keywords: branch-and-bound approach, change point, dynamic programming, remaining cost, segmentation, time series