Features translate raw data into signals models can use. This quiz covers lags, aggregations, leakage, encoding, and normalization patterns that raise predictive power responsibly.
To capture short‑term memory, create lag features and rolling ______ of the target.
statistics (e.g., mean, min/max)
device IMEI
color palettes
CSS grids
Adding future information in training by mistake is called target ______.
leakage
quantization
sparsity
binning
Calendar features such as weekday, month, and ______ often improve demand models.
screen brightness
mouse dpi
image alpha
holiday proximity
When categories are many and rare, consider target/mean encoding with proper ______ to prevent overfit.
regularization and out‑of‑fold scheme
Bezier smoothing
EXIF scrubbing only
palette swaps
Tree models are insensitive to monotonic scaling, but distance‑based models often require ______.
font embedding
CMYK gamut mapping
TIFF export
feature normalization/standardization
For interaction effects, you can add product or ratio features, but ensure they’re derived only from ______ data.
test leakage
rendered pixels
historical/known‑at‑prediction‑time
future labels
Missing values can be informative; one tactic is to include a binary ______ alongside imputed values.
font glyph
image mask color
port number
missingness indicator
For class‑imbalance in churn prediction, pair feature engineering with evaluation beyond accuracy, such as ______.
PSNR
BLEU
SSIM
precision‑recall/PR‑AUC
When using GA4 sources, stitch features with ______ to maintain a stable entity key.
third‑party cookies
hex color codes
User ID or Client ID mapping
GIF metadata
Before deployment, run permutation or SHAP‑style checks to confirm features align with domain logic and not spurious ______.
kernels
sprites
correlations
gradients
Starter
You’re new to feature engineering for predictive accuracy. Revisit key concepts and common pitfalls to build confidence.
Solid
Good grasp of feature engineering for predictive accuracy. Tighten validation, diagnostics, and deployment readiness.
Expert!
Excellent command of feature engineering for predictive accuracy. You’re ready to scale models and mentor others.
