Facoltà di Economia

Aggiornato A.A. 2021-2022

LEARNING OUTCOMES: On completion of the course students will be expected to:

• Have a good understanding of the fundamental issues and challenges of machine learning: data, model selection, model complexity, etc.
• Have an understanding of some popular machine learning approaches.
• Be able to apply statistical concepts to machine learning algorithms
• Be able to design and implement basic machine learning algorithms using the software package R

KNOWLEDGE AND UNDERSTANDING:  The student will understand the main issues related to machine learning. In particular, the problem of the extraction of the relevant information from high dimensional data set will be formalized and explored.

LEARNING SKILLS: Students are expected to learn how to build a theoretical framework and to implement practical solutions for some basic machine learning problems.

PRE-REQUIREMENTS: Machine Learning is a mathematical discipline, and students will benefit from a good background in probability, linear algebra and calculus. Basic programming experience is a plus.

TOPICS

Introduction to machine learning

Basic programming in R

The bias-variance dilemma

Data reduction and signal extraction: theory and methods

Supervised and unsupervised learning

Bootstrap techniques: theory and applications to machine learning

Artificial neural networks

Convolutional neural networks

Genetic algorithms

Support vector machines

TEXTBOOKS:

• Ethem ALPAYDIN, Introduction to Machine Learning, third edition The MIT Press

September 2014

• Karthik, R., and S. Abhishek. "Machine Learning Using R: With Time Series and Industry-Based Use Cases in R." (2019).

Aggiornato A.A. 2020-2021

Updated A.Y. 2020-2021

Overview
The business analytics course is aimed at market-oriented people, who want develop practical skills of how to apply effective analytical models to solve different business-related problems, interpret their results and understand their implications on the business, with the objective of supporting data-driven driven decision-making to optimize the business processes.

Pre-requisites
Basic knowledge of statistics, multivariate descriptive and predictive methods (like regression, principal component analysis and clustering analysis). Also, elementary programming skills are recommended (in any language, although Python is preferred).

Knowledge and understanding
At the end of the course students should be able to understand: (i) how to apply business analytical tools in a supervised and unsupervised approach; (ii) analyze and interpret the models’ outcomes; (iii) extract valuable business-related insights. In particular, students will learn:
• Python programming language;
• Data exploration;
• Data preprocessing;
• Features’ extraction and selection;
• Hierarchical clustering;
• K-means clustering
• Classification measures;
• Imbalanced learning;
• Linear and Logistic regression;
• Decision tree learning;
• Ensemble learning; and
• Cross-validation and hyper-parameters’ tuning.

Applying Knowledge and Understanding
Practical evidence of the concepts will be given with examples using Python programming language, applied on real-world business problems from different industries (food retail, telecommunications, insurance, etc.). To develop the necessary skills, the students will have to practice both in class and at home.

Making Judgements
Students will be able provided a broad range of complex analytical tools and explained in which situations these should be used. On the basis of analytical results they obtain, students will be able to give a business-related interpretation and extract valuable business insights.

Communication Skills
Students will be subject to a constant oral inquiry during the classes. Optionally, the students can apply for the grades’ discussion, which implies an oral defense of their project.

Learning Skills
Students will learn how to solve real-world business problems and to extract valuable business-related knowledge from data to support corporative decision-making processes.

Six weeks of lectures, three classes per week, with the following organization.
Week 1
From business-generated data to value.
• the advantage of data-driven business decisions;
• soft introduction to Machine Learning (ML) through Human Learning (HL): 0/1 classification;
• soft introduction to Data Mining (DM) and Response Models (RM): our practical use case; and
• Python as a tool for Business Analytics.
Introduction to Python (1).
• Introduction to Google Colab;
• Basic data-types;
• Collections;
• Conditional expressions;
• Loops and comprehensions;
• Functions;
• NumPy array: create and manipulate arrays, basic mathematical operations on arrays, stack arrays, index and slice arrays; and
• Summary exercises.
Week 2
Introduction to Python (2).
• Pandas: Series and DataFrame, basic operations, indexing, slicing, conditional selection of axis, summarizing, aggregating, grouping, create, reorder, rename and delete axis; and
• Data visualization with seaborn and matplotlib;
Data preprocessing: how to prepare the data for business analytics?
• Duplicated values;
• Missing values; and
• Possible inconsistencies and “it makes sense” analysis.
Data exploration: getting to know the data and extracting valuable business insights.
• Tabular exploration with pandas; and
• Visual exploration with seaborn and matplotlib;
Week 3
Feature extraction: how to generate valuable features from the data?
• Business-driven feature transformations; and
• Categories’ fusion.
Feature selection: how to select valuable features from the data?
• Features’ worth assessment: scaled mean deviation and chi-squared test for independence; and
• Correlation analysis.
Week 4
Customer profiling: clustering customers’ database.
• Features’ selection for customer profiling;
• Outliers’ removal;
• Hierarchical clustering;
• K-means clustering; and
• Within sum of squares and silhouette analysis; and
• Business-related interpretation of the customers’ segments.
Week 5
Introduction to supervised machine learning:
• Classification measures;
• Imbalanced learning;
• Logistic regression modelling;
• Decision tree modelling; and
• Ensemble-learning: modelling the models; and
• Models’ generalization ability and cross-validation.
Week 6
Customer scoring.
• Features’ selection for customer scoring;
• Response modelling: development of an interpretable model to predict the respondents of a targeted marketing campaign; and
• Interpretation of the analytical outcomes and linkage with business context.
Project support class(es).

Teaching methods
Virtual classroom teaching, discussion of case studies and practical exercising using Python programming language.
Due to the technical restrictions and the remote nature of the classes, the main tool we will use for the didactic purposes is the Google Colab, an intuitive open-source and installation-free tool from Google that requires students to have only a Gmail account and to install the Google Chrome web-browser.

References
- Slides and classes’ videos. The classes’ materials will available and continuously updated in this shared folder: https://drive.google.com/drive/folders/1ild7LWN-EStiXTgLhZyc3C-AikVTh32f?usp=sharing
- Python for Data Analysis: Data Wrangling with Pandas, NumPy, and IPython, by Wes McKinney.
- Hands-On Machine Learning with Scikit-Learn and TensorFlow: Concepts, Tools, and Techniques to Build Intelligent Systems, by Aurélien Géron;
- Introduction to Machine Learning with Python: A Guide for Data Scientists, by Andreas C. Müller and Sarah Guido;
- Fluent Python: Clear, Concise, and Effective Programming, by Luciano Ramalho;
- Data Mining Cookbook: Modeling Data for Marketing, Risk, and Customer Relationship Management (Datawarehousing), by Olivia Parr Rud;
- Business Analytics Using SAS Enterprise Guide and SAS Enterprise Miner: A Beginner's Guide, by Olivia Parr Rud;