CSCI 1109 - M16 - Scaling, encoding, binning, feature creation

This lecture introduces the core pre-processing steps that make later machine learning models actually work: scaling, encoding, binning, and simple feature creation. Using a small credit-scoring style dataset, we look at how mismatched units, naive category encodings, and arbitrary bins can quietly distort distances, decision boundaries, and even fairness. Working only with pandas and NumPy, you’ll practice reshaping raw columns into well-behaved, defensible features while avoiding classic leakage traps so that future models have a solid foundation.

Course module page: https://web.cs.dal.ca/~rudzicz/Teaching/CSCI1109/2026/index.html#module=1109-M16-scaling-encoding-features

Видео CSCI 1109 - M16 - Scaling, encoding, binning, feature creation канала Atlantic AI Institute