Insegnamento a.a. 2025-2026

1. Essential Concepts in Time Series Analysis. Principles of Forecasting and Decision Theory

2. Autoregressive Moving Average (ARMA) Models. Selection and Maximum Likelihood Estimation of AR, MA and ARMA models. Forecasting ARMA processes

3. Unit Roots and the Spurious Regression Problem

4. Univariate Volatility Modeling: Introduction to ARCH and GARCH 

5. Advanced Univariate Volatility Modeling: Non-Gaussian Marginal Innovations; Exogenous (Predetermined) Factors; Forecasting; Estimation and Inference

6. Realize Variance

7. Random Forests and Trees in Financial Applications

8. Support Vector Machines and Regressions in Finance

9. Artificial Neural Networks in Financial Applications

• Explain the fundamental principles of time‐series analysis and forecasting, including decision‐theory concepts, and understand how these principles support empirical research.

• Select, estimate, and forecast AR, MA, and ARMA models using maximum‐likelihood techniques, both conceptually and through optional practical lab exercises.

• Identify and diagnose unit‐root issues and spurious‐regression problems in financial time series, and apply these diagnostic techniques in practical lab sessions.

• Describe and implement ARCH/GARCH frameworks for modeling and forecasting univariate volatility, with hands‐on practice in optional labs.

• Summarize key machine‐learning methods (e.g., random forests, decision trees, neural networks) for empirical finance, and demonstrate their use through optional practical exercises.

• Estimate and interpret ARMA models on real financial time series: given a dataset, select appropriate AR and MA orders using likelihood‐based criteria, fit the model, and critically discuss the estimation output (e.g., parameter significance, residual diagnostics) .

• Conduct unit‐root and spurious regression diagnostics: apply formal tests (e.g., Augmented Dickey–Fuller, Phillips–Perron) to assess stationarity, identify spurious relationships, and draw conclusions about integration properties of the series .

• Fit, validate, and forecast with ARCH/GARCH models: implement an appropriate ARCH or GARCH specification on provided volatility data, check model assumptions (e.g., residual autocorrelation, volatility clustering), and produce out‐of‐sample volatility forecasts .

• Apply tree‐based machine‐learning techniques to asset‐pricing problems: use random forests or decision‐tree algorithms on a given financial dataset to generate out‐of‐sample return predictions, measure forecasting accuracy (e.g., mean squared error, entropy), and interpret variable‐importance outputs .

• Evaluate neural‐network models for empirical finance tasks: construct a basic feedforward neural network to predict a target (e.g., asset return or volatility), assess its performance relative to classical benchmarks, and critically discuss its strengths and limitations in a financial econometrics context .

• Lectures
• Practical Exercises
• Collaborative Works / Assignments

• Lectures: Systematic, instructor-led sessions where the core theoretical content is presented. This includes derivations of time-series properties, decision-theory principles, ARMA/ARCH/GARCH model development, and introductions to machine-learning concepts in finance.
  • During lectures, instructors work through numerical examples (e.g., likelihood-based ARMA estimation, ADF/PP unit-root tests, GARCH fitting) on the board or via live demonstrations, ensuring students grasp both the algebraic derivations and their empirical interpretation.
     

• Practical Labs: Hands-on, problem-solving sessions where students apply theoretical concepts to real data. Exercises focus on model selection, estimation, diagnostic testing, and forecasting.

  •  

    Six optional laboratory sessions are scheduled (announced on the course webpage) in which students work through pre-designed exercises—each corresponding to one or more lecture topics (e.g., fitting ARMA processes, conducting unit-root tests, implementing ARCH/GARCH specifications, running random forests).
  • In each lab, a concrete dataset is provided (e.g., a financial time series with volatility clustering).
     

• Collaborative Works / Assignments: Group-based homework assignments in which teams of 3–4 students tackle an empirical project from start to finish—including data acquisition, model implementation, and written interpretation.

  • Six optional group assignments are due during the semester (tentative schedule by mid-February).

Each of these methods—including lecture presentations, guided practical labs, and collaborative homework—ensures that students can both understand and apply modern time-series and machine-learning techniques in an empirical finance setting.

• Written Exam (approximately 65 % of grade): A single mixed‐format test combining open‐answer questions (interpreting ARMA/ARCH/GARCH output, explaining unit‐root tests, and discussing ML concepts) with closed‐answer items (definitions, properties). This directly measures students’ theoretical understanding and empirical interpretation skills. 
   

• Group Assignments (approximately 35 % of grade): Three team projects (ARMA/unit‐root diagnostics; ARCH/GARCH estimation and validation; ML forecasting) requiring data preparation, model implementation, diagnostic testing, and written discussion. These verify applied econometric and machine‐learning skills, as well as teamwork and communication. 

The material covered is outlined in the lecture slides made available via Blackboard:

https://blackboard.unibocconi.it/

but a firm grasp of the course textbook,

Guidolin, M., Pedio, M., Essentials of Time Series for Financial Applications, 1st Edition, (eBook ISBN: 9780128134108; Paperback ISBN: 9780128134092), Academic Press, May 2018,

is necessary. The exam covers class lectures, slides and the corresponding chapters (if marked with a *) from the book. Any questions concerning whether starred portions of the book are to be covered, will be addressed by copying and pasting this portion of your syllabus.