Use Case: Statistical Upward Propagation of Attributes

Summary

Scope: Plant Layer, Scraper
Level: Scraper Data
Actors: Developer
Brief: Plant attributes can be propagated from concrete plants to higher ranks, if they have a consistent value across all lower ranks
Status: Current
Assignee: Vseslav
Criticality: Important

Precondition:
- Plant attributes and hierarchies have been inserted into the database via scraper.
- A configurable propagation threshold is available.
- The threshold is bigger than 50% (in order to avoid multiple equally frequent qualifying values).
- Empty CSV values are interpreted as null.
Main success scenario:
- The system starts at the lowest taxonomic rank (cultivar).
- For each parent rank and each attribute independently, only non-null descendant values are considered.
- Equal values are grouped.
- For array attributes, arrays are compared as complete values while ignoring element order.
- The most frequently occurring value is determined.
- If its relative frequency among all non-null values reaches the configured threshold, that value is propagated to the parent.
- Propagated values are treated as regular values during subsequent propagation to higher taxonomic ranks.
- The process continues until the highest supported rank has been processed.
Alternative scenario:
- A value does not reach the configured threshold, so no value is propagated.
Error scenario:
- As the use case is implemented at build time (in the scraper), no run-time errors can occur.
Postcondition:
- Higher taxonomic ranks contain propagated attribute values where sufficient consistency exists.
- Original explicit attribute values remain unchanged.
Non-functional Constraints:
- Propagation must be deterministic.
- The propagation threshold must be configurable.

More info on hierarchy: doc/database/hierarchy.md

More info on plant attributes: ../assigned/plant_attributes.md