01. Statistical Learning for Decision Systems
This lecture connects Statistical Learning for Decision Systems to data framing, validation, metrics, monitoring, and decision costs.
Machine Learning Basics for Decision Science
Machine Learning
Data Science
Lecture Notes
Decision Science
This course introduces machine learning as a decision-support discipline. The focus is on designing datasets, labels, splits, models, metrics, and monitoring workflows that can survive operational use.
By the end, a reader should be able to frame supervised and unsupervised learning tasks, control leakage, compare baselines and flexible models, interpret metrics through costs and thresholds, use clustering and dimensionality reduction carefully, and connect model outputs to recommendations.
Lecture Sequence
02. Framing Business Questions as Supervised and Unsupervised Tasks
This lecture develops Framing Business Questions as Supervised and Unsupervised Tasks with examples that make assumptions, diagnostics, and interpretation visible.
03. Data-Generating Processes, Labels, Features, and Representative Data
This lecture uses Data-Generating Processes, Labels, Features, and Representative Data to clarify the analyst’s question, evidence, assumptions, and decision implications.
04. Train, Validation, Test Splits, and Leakage Control
This lecture connects train, validation, and test splits to leakage control and credible performance estimates.
05. Linear and Logistic Models as Predictive Baselines
This lecture frames Linear and Logistic Models as Predictive Baselines as a decision problem and asks what evidence can be trusted, challenged, and communicated.
06. Regularization, Bias-Variance, and Feature Selection
This lecture builds intuition for Regularization, Bias-Variance, and Feature Selection and ties the result to model choice, uncertainty, and action.
07. Tree Models, Random Forests, and Gradient Boosting
This lecture applies Tree Models, Random Forests, and Gradient Boosting with emphasis on diagnostics, tradeoffs, and evidence limits.
08. Regression Metrics, Residuals, and Operational Error Costs
This lecture develops Regression Metrics, Residuals, and Operational Error Costs as a practical pattern for analysis, diagnostics, and decision support.
09. Classification Metrics, Thresholds, and Confusion Matrices
This lecture connects Classification Metrics, Thresholds, and Confusion Matrices to validation, operating points, and decision costs.
10. Probability Calibration and Risk Scoring
This lecture develops Probability Calibration and Risk Scoring with examples that make assumptions, diagnostics, and interpretation visible.
11. Unsupervised Learning for Segmentation and Structure Discovery
This lecture uses Unsupervised Learning for Segmentation and Structure Discovery to clarify the analyst’s question, evidence, assumptions, and decision implications.
12. Clustering Methods: K-Means, Hierarchical, and Density-Based Ideas
This lecture compares clustering methods through structure discovery, stability, and segment usefulness.
13. Dimensionality Reduction, PCA, Embeddings, and Visualization
This lecture frames Dimensionality Reduction, PCA, Embeddings, and Visualization as a decision problem and asks what evidence can be trusted, challenged, and communicated.
14. Feature Engineering for Operational Machine Learning
This lecture builds intuition for Feature Engineering for Operational Machine Learning and ties the result to model choice, uncertainty, and action.
15. Hyperparameter Tuning, Model Comparison, and Pipelines
This lecture applies Hyperparameter Tuning, Model Comparison, and Pipelines with emphasis on diagnostics, tradeoffs, and evidence limits.
16. Error Analysis by Segment and Use Case
This lecture develops Error Analysis by Segment and Use Case as a practical pattern for analysis, diagnostics, and decision support.
17. Monitoring Drift, Degradation, and Feedback Loops
This lecture connects Monitoring Drift, Degradation, and Feedback Loops to validation, governance, and operational response.
18. From Scores, Segments, and Model Outputs to Operational Recommendations
This lecture develops From Scores, Segments, and Model Outputs to Operational Recommendations with examples that make assumptions, diagnostics, and interpretation visible.
19. Fairness, Robustness, and Operational Guardrails
This lecture uses Fairness, Robustness, and Operational Guardrails to clarify the analyst’s question, evidence, assumptions, and decision implications.