| 摘要 |
<P>PROBLEM TO BE SOLVED: To perform the clustering of time by automatically determining the number of clusters when time series data are applied, and automatically estimating and separating true featured values which contribute to the clustering of time and the influence of noise features which do not contribute to the clustering of time. <P>SOLUTION: An input part 1 accepts the input of the time series data of an observation value, and update by a cluster allocation variable update part 23, a feature selection variable update part 24, a feature selection parameter update part 25, a feature observation parameter update part 26, a noise observation parameter update part 27, a cluster transition parameter update part 28, a mixture ratio parameter update part 29, and a hyper parameter update part 210 is repeated until analysis end conditions are satisfied. <P>COPYRIGHT: (C)2013,JPO&INPIT |
