SECOM Baseline Test

Evaluate AI analysis quality using the SECOM dataset for semiconductor manufacturing.

Test AI analysis with the SECOM dataset as a quality baseline.

SECOM Dataset Properties:

  • Rows: 1,567 (semiconductor manufacturing)
  • Columns: 591 (sensor data)
  • Minority Class: 6.6% (defect rate)
  • Time Period: ~64 days

Expected Results:

  • Data Quality Score: 65-95% (variance acceptable)
  • Imbalanced Detection: MUST be detected (6.6% minority)
  • Temporal Coverage: ~64 days
  • Response Time: <60s (typical serverless limit)

Run Test:

  1. Download SECOM.csv from UCI Repository
  2. Upload via your application UI
  3. Select appropriate analysis type (e.g., "Quality Control")
  4. Compare results against baseline

Reference:

The SECOM dataset is available at: https://archive.ics.uci.edu/ml/datasets/SECOM

When to Test:

  • After changes to analysis endpoints
  • After changes to data parsing logic
  • After AI model updates
  • After prompt modifications

Origin

Originally developed for fabrikIQ - AI-powered manufacturing data analysis.