Customer attributes contain many missing values.
Do not replace every missing value with zero without checking meaning.
Choose two appropriate practices.
Investigate why values are missing, compare imputation, exclusion, or indicators, and apply a consistent pipeline in training and inference.
Detailed explanation
Zero may distort the distribution.
The policy and its bias can be evaluated.
Training-serving skew can result.
Missing, not collected, and not applicable can require different treatments.
Check official documentation
Compare the following points with the official documentation.
Review Task 1.3 and the preprocessing and leakage guidance. Distinguish omitted, inapplicable, and failed measurements, then outline how to compare imputation, removal, and missingness indicators and apply the training-fitted processing to validation and production.
Official references
What to confirm
Investigate the meaning of missingness, compare treatment options, and distinguish fitting imputation values from applying the resulting transformation.
Key points
- Missing cause
- Imputation
- Consistency
Notes
- Environment: Reference review of official documentation and hypothetical examples; no AWS resources or model training are required.
- Fit imputation statistics such as means or medians on training data. Applying the same processing to validation does not mean recomputing those statistics with validation data.
- Removal may discard important cases or groups with more missing values. Review the retained population as well as overall performance.
Foundation review
Missing value
A feature value that is absent because it was not recorded, unavailable, invalid, or not applicable.
Missingness indicator
A separate feature recording that an original value was missing.