You work for a manufacturing company that owns a high-value machine which has several machine settings and multiple sensors. A history of the machine's hourly sensor readings and known failure event…

Question

Options

• AData preparation: Daily max value feature engineering; Model training: AutoML classification with
• BData preparation: Daily min value feature engineering; Model training: Logistic regression with
• CData preparation: Rolling average feature engineering; Model training: Logistic regression with
• DData preparation: Rolling average feature engineering; Model training: Logistic regression with

• A
  5% (2)
• B
  10% (4)
• C
  2% (1)
• D
  83% (34)

Explanation

Rolling average feature engineering is the best choice for sensor time-series data because it smooths out short-term noise while preserving meaningful trends leading up to a failure event, providing a richer temporal signal than a single daily max or min value. Logistic regression is the natural model choice here because the target label is binary - the machine either fails within 3 days or it does not. It is interpretable, computationally efficient, and well-suited for this type of binary classification problem. AutoML classification (Option A) could also work but is a heavier solution for a task logistic regression handles well. Daily max/min (Options A and B) captures only extreme values and discards the temporal trend information that rolling averages preserve.

No community discussion yet for this question.