Feature Engineering for Machine Learning: Encoding, Scaling, Binning & Leakage

About this course

A feature engineering course that prepares a synthetic loan-default dataset for one fictional online lender, Brightfield Lending, in a spreadsheet, for data science students, analysts moving into machine learning, ML engineers who want to see every step and interview candidates. You find missing incomes, add a missing-indicator feature and impute with the median and group means, turn categories into one-hot dummies, count encodings and ordinal ranks without the dummy-variable trap, fit min-max, z-score and robust scalers on the training rows only, cut tenure and income into bands and quantile buckets with default rates, log-transform and winsorise a skewed income, build loan-to-income, DTI and interaction features, compare naive, train-only and smoothed target encoding to see why leakage happens, and screen features with correlation and a univariate AUC. Synthetic data for learning. The first lab is free.