Projects |

Link Prediction in Citation Networks

Tue, 01 Feb 2022 00:00:00 +0000

This project presents our solution for the CentraleSupélec Data Science Kaggle Competition, where we achieved 2nd Place for the task of link prediction in citation networks.

The goal of the project was to predict missing links between papers in a citation graph using machine learning and graph-based methods. The task required understanding the structure of citation networks and designing models that could effectively infer whether an edge should exist between two nodes.

Our approach focused on leveraging graph representations and learning meaningful relationships between papers based on the network structure. This project highlights the use of Graph Neural Networks (GNNs) and related techniques for relational learning on graph data.

You can view the full winning solution here:

Twitter Sentiment Analysis on Stocks

Tue, 01 Jun 2021 00:00:00 +0000

This project analyzes Twitter sentiment related to stock market discussions to understand how social media opinions correlate with stock movements.

The goal of the project was to extract insights from large volumes of Twitter data and classify tweets based on their sentiment toward specific stocks. By applying Natural Language Processing (NLP) and machine learning techniques, the model identifies whether tweets express positive, negative, or neutral sentiment.

The workflow includes data preprocessing, feature extraction from text, sentiment classification, and analysis of sentiment trends related to financial markets. This project demonstrates how social media data can be leveraged for financial analytics and market sentiment understanding.

You can explore the project repository here: