hard20 pts · first answer countsmachine-learningfeature-engineeringdata-leakage
A rolling mean that peeks ahead
A feature is built with demand.rolling(7, center=True).mean(). Offline accuracy is excellent but live accuracy is poor. Why?
Discussion 0
Join free to post in the discussion.
Related questions
- Which model needs feature scalingMachine Learning Engineer
- Encoding a city featureMachine Learning Engineer
- Predicting a flat price in rupeesMachine Learning Engineer
- One gradient descent stepMachine Learning Engineer
- Sigmoid at zeroMachine Learning Engineer
- Logistic regression with five classesMachine Learning Engineer
- The naive part of Naive BayesMachine Learning Engineer
- Entropy of a 50/50 nodeMachine Learning Engineer