Learning objectives

The goal of the course is to provide students with mathematical, statistical and computational tools for a rigorous understanding of machine learning. A central aspect is the critical discussion of how and to which extent machine learning methods are essential in large scale data analysis in order to develop a professional profile combining solid quantitative skills with an in-depth knowledge of economic and corporate dynamics to support strategic decisions based on data analysis. At the end of the course students will be able to master classical methods of machine learning, implement data analysis algorithms, choose the most suitable techniques, identify relevant structures underlying the data for prediction purposes, critically discuss the output generated by a machine learning technique.

Prerequisites and basic notions

A working knowledge of statistics as covered in basic statistics and econometrics courses is assumed.

Program

• Introduction to Statistical Learning
• Linear Regression Models and Least Squares
• Gauss-Markov Theorem
• Best Subset Selection
• Shrinkage Methods
• Ridge Regression
• Lasso Regression
• Linear Classification Methods
• Bayes Classifier
• Linear Discriminant Analysis
• Logistic Regression
• Model Evaluation & Selection
• Bias-Variance Tradeoff and Model Complexity
• Cross-Validation
• Neural Networks Introduction
• Neural Network Training
• Clustering Methods

Bibliography

Didactic methods

The course consists of 36 contact hours: 24 hours of lectures (equivalent to 4 ECTS credits) and 12 hours of lab sessions (equivalent to 2 ECTS credits)

Learning assessment procedures

The exam will assess:
(a) The understanding of theoretical tools (concepts and formal models) presented in the course,
(b) The ability to use these theoretical tools to discuss the results of data set analysis.
At the end of the course in April, an assignment will be issued, consisting of a theoretical part and a computational part. Approximately one month will be given for its completion. Sample examinations from past academic years will be made available on the course's Moodle site. If a grade of 18/30 or higher is achieved on the assignment, it will remain valid for the entire summer exam session (until July 2026). The student may then choose in which exam sitting (first or second) to take the corresponding oral test. This oral examination will consist of a discussion and preliminary review of both the theoretical and computational parts, followed by an oral test focusing solely on the theoretical part, lasting approximately 30 to 40 minutes. This exam can be taken only once. Students who fail to submit or pass the assignment, or who fail the corresponding oral exam, must take a comprehensive oral test covering both the theoretical and computational parts of the course, including solving exercises on the spot.

Evaluation criteria

Written Exam Components & Weighting:
Theoretical portion: 2/3 of total grade,
Software application portion: 1/3 of total grade

Criteria for the composition of the final grade

The final examination grade is determined by computing the arithmetic average of the written and oral component scores.

Exam language

Inglese