EDA

Missing Values

1. Completely at random (collection system issue?)

2. At random (people not entering their info)

3. Not at random (happening for a reason)

Noise Techniques

Smoothing

• Separate into bins, replace with bin mean or with nearest bin boundary

• Regression line

• Distribution line

Discretization

• Helpful for concept hierarchies, think age 67 vs 68

Normalization / Standardization