Simple anomaly detection

lfoppiano / supercon2

Staging-area for automatically collected experimental data for the SuperCon database with a curation interface with enhanced-document viewer and curation-ready interface

https://supercon2.readthedocs.io

4 stars 0 forks source link

Simple anomaly detection #180

Closed lfoppiano closed 1 year ago

lfoppiano commented 1 year ago

Implement a simple anomaly detection to identify records with suspicious values

test only "new" records
check Tc or Formula (might be too strict):
- if Tc can be parsed and is higher than 270 K
- if the formula exists, there are no variables, and cannot be parsed (might be too strict)
if anomaly:
- set record as "invalid"
- set error_type as "anomaly_detection"
- generate training data

We can add few more rules on this

lfoppiano commented 1 year ago

After modification we have a new error_type