--interpolate_aggregated_feats

Switch

--interpolate_aggregated_feats DAYS

Description

Aggregates message-level features up to a new group-by field, interpolating across the given number of days for groups with sparse/missing dates. Aliased as --interpolate_feats.

Argument and Default Value

A float — the number of days to interpolate across. Default: None.

Details

Verified requirement: the feature table given by -f must be at the message level (grouped by your --messageid_field, not e.g. user_id) — FeatureRefiner.createInterpolatedFeatTable() asserts this explicitly and fails with AssertionError: Interpolate currently only works if interpolating message-level features otherwise.

Known issue — broken in SQLite mode. Once past that check, it calls mm.getTableColumnNameTypes(...), which queries information_schema.columns — not supported by SQLite. Verified: fails with sqlite3.OperationalError: no such table: information_schema.columns. Works against MySQL.

Other Switches

Required Switches:

• -d, -g, -t

• -f (a message-level feature table)

• --date_field

Example Commands

Requires MySQL — broken against SQLite, see Known issue above:

dlatkInterface.py -d dla_tutorial -t msgs -c user_id -f 'feat$1gram$msgs$message_id' \
--date_field date --interpolate_aggregated_feats 3