Using fuzzy neural network clustering algorithm in the symbolization of time series

Data mining on time series needs to translate the continuous time series into discrete symbol sequences first. In this paper, a new and efficient approach to convert the time series into symbol sequence is proposed. In the approach, the time series is converted into a discrete sequence with a piecewise linear segmentation representation first, each segment has a simple and primitive shape; then, the segments are clustered by using a fuzzy neural network clustering algorithm. The clustering is based on a similarity measure that can describe the shape similarity of vectors. Results of experiment show that the fuzzy neural network and the shape similarity measure are suitable to the online clustering analysis of time series