A Spacecraft Electrical Characteristics Multi-Label Classification Method Based on Off-Line FCM Clustering and On-Line WPSVM
Fig 2
Flow chart of the off-line system.
The off-line system gets a large number of signal data as input. These signal data are first processed by the pre-process procedure. The pre-process aims to divide a time series into several events by threshold value. When the time series traversal is over, the off-line system will put all of these samples into the fuzzy c-means section. The fuzzy c-means clustering is a process to produce the training set. This process clusters similar waveform data together and labels those waveforms to find the same category, then generates a clustering center for every class.