Optimized fuzzy information granulation of temporal data

It accords with the human´s intelligence for data processing that we do information granulation of time series and then study and analyze the time series on the different level of granularities. There are two kinds of traditional methods for the information granulation of temporal data. One implements the granulation by first dividing the time series into segments in terms of the granularity of the time-axis, and then building granules on each segment, without taking into account the distribution information on the value-axis. The other implements the granulation in terms of the granularity of value-axis, ignoring the information of time granularity adopted by humankind. This paper discusses the fuzzy information granulation of time series based on the optimization of time granularity and value granularity. In detail, we improve original dynamic time warping method, and propose a new method to search for the ideal result of information granulation of time series based on time-axis if we consider the value information. It is achieved through calculating the similarity of two results of information granulation of one time series based on both time-axis and value-axis. Thus, we build up a good platform for the granular analysis of time series. Experiment results show that the optimized fuzzy information granulation of time series is of good performance.