Machine Learning in Finance Course

Lesson 1 • Core ML Concepts for Finance

Introduces supervised, unsupervised, and reinforcement learning paradigms. Connects each paradigm to recurring financial tasks such as prediction and clustering.

Lesson 2 • Python Environment for Financial ML

Sets up a reproducible Python stack with key libraries. Students run end-to-end data pipelines before tackling modelling chapters.

Lesson 3 • Financial Data Types and Sources

Surveys structured and alternative data used in finance. Students learn to assess data quality, frequency, and licensing before modelling.

Lesson 4 • Statistical Foundations for Financial ML

Reviews probability, distributions, and hypothesis testing as applied to financial series. Provides the statistical toolkit required for model evaluation later.

Lesson 1 • Labelling and Target Construction

Defines forward-looking labels for classification and regression tasks. Proper labelling prevents look-ahead bias and aligns targets with trading objectives.

Lesson 2 • Time-Series Cross-Validation

Implements purged and embargo cross-validation for financial data. Prevents information leakage across training and validation folds.

Lesson 3 • Data Cleaning and Quality Control

Addresses missing values, outliers, and corporate actions in price data. Clean data is the prerequisite for all feature engineering steps that follow.

Lesson 4 • Feature Engineering for Price Series

Derives technical and statistical features from OHLCV data. These features form the primary input space for predictive models in later chapters.

Lesson 1 • Hyperparameter Optimisation

Applies grid search, random search, and Bayesian optimisation to financial models. Efficient tuning reduces computational cost while improving out-of-sample performance.

Lesson 2 • Support Vector Machines and Kernels

Applies SVMs to financial classification with linear and nonlinear kernels. Margin maximisation provides theoretical guarantees useful in noisy markets.

Lesson 3 • Tree-Based Models for Classification

Introduces decision trees, random forests, and gradient boosting for directional prediction. Ensemble methods improve robustness over single-model approaches.

Lesson 4 • Model Evaluation and Selection

Defines financial-specific metrics beyond accuracy, including Sharpe-based scores. Students select models using statistically sound comparison frameworks.

Lesson 1 • Anomaly Detection in Financial Data

Detects outliers and regime shifts using isolation forests and autoencoders. Anomaly signals feed into risk management and fraud detection workflows.

Lesson 2 • Dimensionality Reduction Techniques

Covers PCA, t-SNE, and UMAP for compressing high-dimensional financial features. Reduced representations improve model stability and visualisation.

Lesson 3 • Latent Factor Models

Extracts latent risk factors using factor analysis and NMF. Connects data-driven factors to traditional asset pricing theory.

Lesson 4 • Clustering for Asset Grouping

Applies k-means, hierarchical, and DBSCAN clustering to group assets by behaviour. Clusters inform diversification and sector-agnostic portfolio construction.

Lesson 1 • Recurrent Networks for Time Series

Implements vanilla RNNs and LSTMs on price and volume sequences. Gating mechanisms address vanishing gradients in long financial histories.

Lesson 2 • Neural Network Fundamentals

Covers feedforward networks, activation functions, and backpropagation. Provides the architectural foundation required before introducing recurrent models.

Lesson 3 • Transformer Models in Finance

Introduces attention mechanisms and Transformer encoders for financial sequences. Self-attention captures non-local dependencies across long time horizons.

Lesson 4 • Training and Regularisation Strategies

Addresses overfitting, learning rate scheduling, and early stopping for financial neural nets. Proper training protocols are critical given limited non-stationary data.

Lesson 1 • Portfolio Backtesting Framework

Builds a rigorous backtesting engine with realistic assumptions. Students identify and correct common backtest biases before live deployment.

Lesson 2 • Alpha Signal Generation and Combination

Converts ML model outputs into tradable alpha signals. Signal combination via stacking and blending improves information ratio.

Lesson 3 • Mean-Variance and Beyond

Applies Markowitz optimisation and its extensions using ML-estimated inputs. Addresses estimation error in covariance matrices with shrinkage methods.

Lesson 4 • Transaction Costs and Turnover Control

Incorporates realistic cost models into portfolio optimisation. Turnover constraints prevent alpha erosion from excessive rebalancing.

Lesson 1 • Volatility Forecasting Models

Compares GARCH, realised volatility, and LSTM-based forecasts. Accurate volatility estimates feed directly into options pricing and risk limits.

Lesson 2 • Market Risk and Value at Risk

Estimates VaR and expected shortfall using historical simulation and ML-enhanced methods. Tail risk modelling improves on parametric normality assumptions.

Lesson 3 • Fraud and Anomaly Detection

Detects fraudulent transactions and market manipulation using supervised and unsupervised ML. Imbalanced class handling is central to effective fraud models.

Lesson 4 • Stress Testing and Scenario Analysis

Generates adversarial scenarios using GANs and historical analogs. Stress tests reveal portfolio vulnerabilities beyond standard VaR estimates.

Lesson 1 • Model Governance and Audit Trails

Establishes documentation, validation, and approval workflows for financial ML models. Governance frameworks satisfy regulatory expectations for model risk management.

Lesson 2 • Real-Time Inference and Latency

Optimises models for low-latency scoring in trading and risk systems. Latency budgets differ across use cases from high-frequency to end-of-day.

Lesson 3 • Model Monitoring and Drift Detection

Tracks data drift, concept drift, and performance degradation in production. Early detection prevents silent model failures in live financial systems.

Lesson 4 • MLOps for Financial Applications

Covers CI/CD pipelines, model registries, and containerisation for financial ML. Automated pipelines reduce deployment risk and accelerate iteration cycles.