
Already know some of this? A 2-minute test skips the lessons you know.
Take the placement testA free 14-lesson course built from freeCodeCamp.org's 2h 09m of YouTube video. 2 days at 90 minutes a day, with study notes and a quiz in every lesson. A 2-minute placement test skips what you already know.
Taught by freeCodeCamp.org on YouTube. Videos play in YouTube's own player, so the creator keeps every view.
Understand the course structure, scikit-learn's estimator API (fit/predict), and how to build basic pipelines with preprocessing and models.
Learn why evaluating on training data is misleading and how cross-validation with GridSearchCV provides robust hyperparameter tuning.
Recognize ethical issues in datasets (e.g., racial features in Boston housing) and the responsibility to inspect data before modeling.
Apply StandardScaler for feature scaling and understand its sensitivity to outliers through visual demonstrations.
Use QuantileTransformer to scale data robustly against outliers by leveraging quantiles instead of mean/variance.
Generate polynomial features to enable linear models to capture non-linear patterns in classification tasks.
Encode categorical variables using OneHotEncoder and handle unseen categories during inference with handle_unknown='ignore'.
Compute precision, recall, and F1 scores; understand their trade-offs and how to use them in GridSearchCV via make_scorer.
Create a custom scorer that optimizes for the minimum of precision and recall to balance both metrics during hyperparameter search.
Incorporate sample weights (e.g., transaction amounts) into custom metrics to prioritize high-impact predictions in fraud detection.
Use IsolationForest for unsupervised outlier detection and adapt its outputs to scikit-learn compatible classification metrics.
Wrap domain-knowledge rules (e.g., fare thresholds for Titanic survival) into FunctionClassifier for grid-searchable, pipeline-compatible models.
Draw polygons interactively to define classification regions, export as JSON, and deploy as InteractiveClassifier with robustness to missing data.
Use drawn polygons for outlier detection and build fallback classifiers that defer to rules when model confidence is low or outliers detected.