In this step, we have our candidates generated from the original dataset.
We must now validate these candidates. To do this, we will use as a restriction the occurrences. There are two types of occurrences in our approach : - Global occurrences (GO)
: the number of occurrences in the entire dataset - Spatial occurrences (SO)
: the number of spatial-series containing the candidate
The thresholds are used to select from the list of candidates the real motifs
Sigma
: threshold to check the minimal number of global occurrences of each motif.Kappa
: threshold to check the minimal number of spatial occurrences of each motif.Here an example with three scenarios:
Example of encoded spatial-time series
Result of the three scenarios
Filter motifs by removing isolated motifs and rechecking occurrences.
To find these isolated motifs we use the cluster method.
Cluster method
If the motif does not have a neighbor then it is deleted. After that when all isolated motifs are deleted, we recheck like the previous step the occurrences.