Skip to content

Use Case 1: Chips

Method 1: Break

  • Chips predicted quality vs real quality (classification)
  • Model test on new data (not previously seen by model)
  • Ensemble model from different analysis windows
  • Overall accuracy: 83%
  • Dataset:
    • Method: Break
    • Speed: 0.2mm/s

Method 2: Stacked

  • Chips predicted quality vs real quality (classification)
  • Model test on new data (not previously seen by model)
  • First model: Finetuning possible
    • Feature selection
    • Combining different models from different analysis windows
  • Overall accuracy: 93%
  • Dataset:
    • Method: STACKED
    • Speed: 2mm/s