Machine Learning

Semester 2 · 27503 · Master in Data Analytics for Economics and Management · 6CP · EN

This course offers a comprehensive introduction to the core concepts, techniques, and algorithms of machine learning, as well as some platforms commonly used in practice. Students will explore essential topics such as data preprocessing—including data manipulation, transformation, feature selection, and dimensionality reduction— followed by key methods in supervised learning like regression and classification. The course covers unsupervised learning approaches such as clustering and association rule mining. Moreover, Artificial Neural Networks are covered through the study of the perceptron, the multi-layer perceptron. An overview of deep networks and multi-task deep learning is provided. Foundational ideas, principles and applications of Reinforcement Learning are also covered. Throughout the course, students will not only develop a solid understanding of the theoretical underpinnings of these algorithms but also acquire practical skills in implementing data workflows, applying machine learning methods to real-world data, and evaluating model performance. Applications across diverse domains are discussed to illustrate the impact and versatility of machine learning.

Lecturers: Andrea Rosani, Giuseppe Di Fatta

Teaching Hours: 40
Lab Hours: 20
Mandatory Attendance: The attendance is not compulsory, but students are highly encouraged to attend both lectures and labs.

Course Topics
The main topics include: • Data Analysis • Model selection • Unsupervised learning • Supervised learning • Deep learning • Reinforcement learning

Teaching format
Frontal lectures, lab assignments, project work.

Educational objectives
Knowledge and understanding: The student acquires programming knowledge especially aimed at data analysis and statistical methodologies for the implementation of models as well as the analysis of large datasets. In particular, computer knowledge is oriented towards machine learning methods, knowledge of modern data management and storage techniques, also from heterogeneous sources in terms of type and structure, including spatio-temporal data and high-dimensional data also in a cloud environment, and the implementation of algorithms for massive data. This knowledge is then complemented by the indispensable knowledge of textual data and network analysis and aspects related to the security and privacy of such data. Applying knowledge and understanding: Ability to apply and implement, through the development of algorithms, techniques for the analysis of large datasets and for spatial and temporal data, under conditions of uncertainty, in order to guarantee the usefulness, quality and effectiveness of the analysis. Capacity to use IT technologies, techniques and methodologies for acquiring, managing, integrating, analysing and visualising large datasets, in order to ensure scalability with respect to the volume and speed of data acquisition. These skills concern in particular database management systems and large datasets and related visualisation techniques, models and languages for expressing data semantics, learning techniques, decision-making models, organisation of information systems, web search techniques, data flow management techniques. Making judgements: Master graduates will have the ability to apply the acquired knowledge to interpret data in order to make directional and operational decisions in a business context. Master's graduates will have the ability to apply the acquired knowledge to support processes related to production, management and risk promotion activities and investment choices through the organisation, analysis and interpretation of complex databases. Communication skills: Master's graduates will be able to communicate effectively in oral and written form the specialised contents of the individual disciplines, using different registers, depending on the recipients and the communicative and didactic purposes, and to evaluate the formative effects of their communication. Learning skills: "MSc graduates should be familiar with the tools of scientific research. They will also be able to make autonomous use of information technologies to carry out bibliographic research and investigations both for their own training and for further education. In addition, through the curricular teaching and the activities related to the preparation of the final thesis, they will be able to acquire the ability - to identify thematic links and to establish relationships between methods of analysis and application contexts; - to frame a new problem in a systematic manner and to implement appropriate analysis solutions; - to formulate general statistical-econometric models from the phenomena studied.

Assessment
• A project, which consists in applying/implementing machine learning algorithms to real-world data, describing the approach and the adopted solution, and presenting the results of an experimental analysis. • A final oral exam with questions on the content of the course.

Evaluation criteria
• Project: 50% of the final mark • Oral exam: 50% of the final mark Note: both project and exam are required to be passed. Criteria for awarding marks Oral exam: ability to present and explain machine learning concepts, methods and algorithms. Ability to select appropriate solutions for machine learning problems. Project: ability to implement data workflow to apply machine learning algorithms to real-world problems, correctness and clarity of the solution, experimental results, ability to solve machine learning problems with the appropriate technique.

Required readings

· Introduction to Data Mining, by Pan-Ning Tang, M. Steinbach, A. Karpatne, V. Kumar. Pearson Education Ltd (2nd Edition, 2020).

Download as pdf

Sustainable Development Goals
This teaching activity contributes to the achievement of the following Sustainable Development Goals.