Statistical learning, classical ML algorithms, regression, classification, clustering, and evaluation metrics.
Understand the four fundamental paradigms of Machine Learning: Supervised, Unsupervised, Semi-Supervised, and Reinforcement Learning with real-world analogies and algorithm taxonomy.
Comprehensive deep-dive into Regression and Classification, decision boundaries, bias-variance tradeoff, algorithm selection matrix, and end-to-end ML workflows.
Step-by-step practical guide: synthetic dataset generation, train/test splitting, StandardScaler normalization, LinearRegression, metrics (MAE, RMSE, R²), and model serialization with Joblib.
End-to-end classification pipeline: Breast Cancer diagnostic dataset, stratified splitting, LogisticRegression, Confusion Matrix, Precision-Recall, ROC-AUC curves, and deployment inference.
Mathematical induction and strategic selection for regression (MAE, MSE, RMSE, R², MAPE) and classification metrics (Accuracy, Precision, Recall, F1-Score, ROC-AUC curves).
Master robust model evaluation and optimization: K-Fold cross-validation, GridSearchCV, RandomizedSearchCV, Bayesian optimization principles, and leak-proof Scikit-Learn pipelines.
Master data preprocessing pipelines and class imbalance mitigation: ColumnTransformer imputation, One-Hot encoding, standard scaling, and leak-proof SMOTE resampling inside cross-validation.
Master advanced ensemble paradigms: Bagging vs Boosting mathematical principles, XGBoost, LightGBM leaf-wise training, and out-of-fold Stacking & Blending meta-learners.
Understand why machine learning models make predictions: Global vs Local interpretability, Permutation Importance, Partial Dependence Plots, LIME surrogate explanations, and Game-Theoretic SHAP values.
Bridge the gap from notebook to production: Pipeline serialization with Joblib, high-performance REST API serving with FastAPI and Pydantic, Docker containerization, and data/concept drift monitoring.
Comprehensive technical interview guide: Supervised learning taxonomy, algorithm derivations, metric selection matrices, overfitting remedies, data leakage prevention, and real-world system scenario questions.