Digital Assets · KAEL Editorial Department
Time in market data: an easily overlooked variable
In automation research, time is not a subsidiary field. The difference between event occurrence, data arrival, and system processing affects the validity of the experiment.
Keep multiple timestamps
Record the source event time and local reception time respectively. During network delays or batch supplementary recording, the two may be significantly different, and mixing them will affect data sorting and feature construction.
unified time base
Use clear time zones and precision, and state cross-source alignment rules. When aggregating by minutes, boundary attribution must also be agreed upon to prevent the same event from being placed in different intervals.
Check lateness data
Clarify whether late records will be compensated, and whether the saved results will be changed after the supplementary calculation. Historical research and real-time processes should share the same set of rules as much as possible to reduce environmental differences.
