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 2 types of occurrences in our approach : Global occurrences (GO)
and Spatial occurrences (SO)
There is possibility to set the filtering with 2 parameters :
Sigma
: threshold of the global occurrencesKappa
: threshold of the spatial occurrencesHere an example with 3 scenarios :
Example of encoded spatial-time series
Result of the 3 scenarios
Filter motifs by removing isolated motifs and rechecking occurrences.
To found 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.