Machine Learning

1What is machine learning?Learned rules instead of written rules: supervised, unsupervised and reinforcement learning, and when ML is the wrong tool.8 min 2Train/test split & your first modelHold out data honestly, train a scikit-learn classifier, compare it against a baseline and read the report.7 min 3Linear & logistic regression intuitionHow linear models draw lines and curves, what coefficients mean, and when to reach for them first.7 min 4Pipelines & preprocessingWrap imputing, scaling and encoding with your model in one Pipeline so training and production never drift apart.6 min 5Evaluating modelsPick metrics that match the decision: confusion matrix, precision, recall, ROC-AUC, and MAE, RMSE and R-squared for regression.7 min 6Decision trees, random forests & gradient boostingFrom a single if-else tree to bagged forests and boosted ensembles that win on tabular data.7 min 7Unsupervised learning: k-means clustering & PCAFind customer segments without labels and compress many features into a few components.7 min 8Hyperparameter tuning, cross-validation & imbalanced dataTune models honestly with k-fold CV and search, and handle rare classes like fraud or churn properly.6 min 9Project: churn model saved & served with joblibEnd-to-end telecom churn project: features, pipeline, evaluation, joblib persistence and a Django endpoint.7 min