Analytics/Data Lake/ORES/Historified scores

From Wikitech
Jump to navigation Jump to search


Column Type Comment
rev_id bigint Revision ID
rev_timestamp string Timestamp when revision was created
revision_parent_id bigint Revision parent ID
user_id bigint User ID creating this revision
user_text string User name creating this revision
model_version string ORES model version
prediction array< string > The most probable class predicted.  Note that many use cases should do a threshold calculation instead.  Some models can emit zero or multiple predictions, others will always emit just one.
probability array< struct< name:string, value:double > > Predicted probability for each class
page_id bigint MediaWiki page ID where this revision appears
page_namespace int MediaWiki page namespace ID
page_title string Current title of MediaWiki page edited
snapshot string Snapshot YYYY-MM corresponding to a mediawiki_history dataset
wiki string Wiki database name
model string ORES model name