An Optimized Bottom-Up Time Series Segmentation Algorithm
Wei Suhua · Journal of Shenyang Jianzhu University · 2007
Segmentation is the basis of the time series data mining,its advantages directly influence the analysis' result.In order to evaluate segmentation algorithm's quality correctly,we proposed a new segmentation evaluation criterion that an algorithm which has the smaller sum of error under the same compression is better.The new criterion considered both of the compression and the fidelity.We also introduce a novel optimized Bottom-Up time series segmentation algorithm.The new algorithm removes the classic Bottom-Up algorithm's even limitation.Experiments on the real-life datasets show that the new criterion is more reasonable than the existing criteria and the novel optimized Bottom-Up algorithm is superior to the classic one.