Research

2024-2025

These are the projects we will start in the first semester of 2024-2025:

• Financial News Sentiment Analysis: Develop a machine learning model that analyzes headlines or summaries of financial news, providing a sentiment score to assess market mood and predict potential market movements.

• Instagram AI: Create an AI for Instagram that generates posts, captions, and stories based on selected topics. The AI will continuously learn and adapt to improve engagement, aiming to increase follower count.

• Match Result Prediction Algorithm: Using machine learning, this project predicts football match outcomes, drawing on current form and historical trends. Additionally, it attempts to forecast the final standings of the Italian Serie A season.

• Company Bankruptcy Prediction: Develop a model that predicts company bankruptcy based on financial indicators such as ROA, ROI, and investment round results. This project seeks to pinpoint the key metrics that signal financial distress.

• BRAILLE Detection: Use machine learning to accurately detect Braille letters from images, contributing to accessibility tools and technology.

• Reinforcement Learning (RISK or Monopoly): Apply reinforcement learning to discover optimal strategies for winning games like RISK or Monopoly. The project analyzes which factors have the greatest impact on a player's chances of success.

• Pneumonia Image Classifier: Build an algorithm that detects pneumonia in chest X-ray images, leveraging a dataset of patients with and without the condition to aid in early diagnosis.

• YouTube Comment Generator and Popularity Analysis: Train machine learning algorithms, such as RNNs, to generate YouTube comments. Additionally, analyze which factors affect a video's popularity and success.

• SMS Spam Detection: Create an algorithm capable of identifying and classifying spam messages, improving communication security.

• AI ASSISTANT: Develop an AI Assistant powered by Retrieval-Augmented Generation (RAG) technology that will serve as a practical tool for Bocconi students, aiding them in various aspects of their daily lives, from academic support to personal productivity. We will efficiently gather and process data from Alumni and student interviews, podcasts, and web scrape official Bocconi sources to set up a unique Knowledge Base. Following the MVP phase, the project aims to evolve into a startup and participate in acceleration programs. Click here for more info

• MORE...

2023-2024

During the second semester of the 23/24 academic year, we worked on the following projects:

• Harmful Brain Activity Classification: Detect and classify seizures and other types of harmful brain activity. You will develop a model trained on electroencephalography (EEG) signals recorded from critically ill hospital patients.

• Digit Recognizer: Correctly identify digits from a dataset of tens of thousands of handwritten images.

• Connect X: In this game, your objective is to get a certain number of your checkers in a row horizontally, vertically, or diagonally on the game board before your opponent. When it's your turn, you “drop” one of your checkers into one of the columns at the top of the board. Then, let your opponent take their turn. This means each move may be trying to either win for you, or trying to stop your opponent from winning.

• Reinforcement learning: Use a double dueling DQN to train an agent that can achieve a superhuman level at the famous Atari Breakout game ("ALE/Breakout-v5"). The observations are images. To simplify the task, you should convert them to grayscale (i.e., average over the channels axis) then crop them and downsample them, so they’re just large enough to play, but not more. An individual image does not tell you which way the ball and the paddles are going, so you should merge two or three consecutive images to form each state. Lastly, the DQN should be composed mostly of convolutional layers.

• Recommendation algorithms: using Movielens 10m data set, our goal is to predict the rating an user would give to a movie she haven’t watched yet. We’ll explore matrix factorization algorithms and some computational statistical methods such as GLMMs. Furthermore, we’ll use Bayesian methods for the matrix factorization and will implement hierarchical Poisson factorization using nonparametric Bayesian methods.

• Urban Housing Price Cluster Analysis: This project focuses on analyzing housing price data across city neighborhoods or entire countries to identify and understand price clusters. By examining historical data, we aim to track how different areas have evolved in terms of affordability, highlighting trends in urban development and economic shifts. This streamlined analysis offers insights into the changing dynamics of real estate values over time.

During the first semester of the 23/24 academic year, we worked on the following projects:

• Credit Card Fraud Detection: Address the critical task of detecting fraudulent credit card transactions. Contribute to the security of financial institutions and cardholders. presentation link

• Loan Default Prediction: Help financial institutions make informed decisions by predicting loan defaults. Ideal for those interested in risk assessment and credit modeling. presentation link

• Starbucks Review Sentiment Analysis and Rating Prediction: Analyze Starbucks reviews, predict ratings, and explore geospatial trends. Perfect for those interested in natural language processing and data-driven customer insights. presentation link

• Heart Failure Prediction: Join the mission to predict heart failure and make a positive impact in healthcare. No medical background required; just a passion for data-driven solutions. presentation link

• E-Commerce Customer Segmentation and Customer Lifetime Value Analysis: Unlock the secrets of customer behavior and drive business growth. Learn valuable skills in customer analytics and segmentation. presentation link

• Applying Convolutional Neural Networks (CNNs) to Electrical Circuit Analysis, Fault Detection, and Load Forecasting in Collaboration with Students from KU Leuven: Implement full electrical circuit analysis, fault detection, and load forecasting using CNNs. Collaborate with students from KU Leuven to transform electrical engineering practices. (in progress)

• Image Classification with CNN: Explore deep learning and CNNs to create accurate image classification models. Get hands-on experience with image data and cutting-edge technology. presentation link

• ETL Project with Scavenger AI:  The project is a collaborative effort with Scavenger AI to build an ETL (Extract, Transform, Load) pipeline. This initiative aims to leverage the expertise of our team to efficiently extract data from diverse sources, transform it into a consistent format, and load it into a designated system for further analysis and insights. This partnership is focused on enhancing data processing capabilities and achieving more accurate and actionable outcomes from the data handled.