Entity Metrics#
Entity-level metrics measure the distance between single point-in-time observations within sequences.
The implemented metrics are based on the comparison of the entity’s features and does not handle the timestamp of an entity. In the spirit of TanaT the timestamps are handle at the sequence level. More specifically, an entity metric can be apply whatever the nature of the entity (event, state or interval).
Sequence-level metrics rely on an entity metric as a building block.
Note
Hamming is the most common choice for categorical features while L2 metrics is most suitable with numerical features. The combined metric enables to define a metric based several features (numerical or categorical) through a linear combinaison of other entity metrics.
