Social Network Analysis (SNA) Teaching Toolkit

This project is designed to teach Social Network Analysis (SNA) to students in the Engineering and Policy Analysis (EPA) MSc program at TU Delft. By leveraging interactive and practical tools, the project aims to provide students with a deep understanding of SNA concepts, methods, and their applications in real-world scenarios.

Objectives:

• Introduce students to the fundamental principles of social network analysis.

• Provide hands-on experience with SNA tools and techniques through Jupyter Notebooks and interactive examples.

• Enhance students’ ability to analyze and visualize complex networks in engineering and policy contexts.

Computational Rumour Detection without Non-Rumour Data

This project introduces a novel approach to rumour detection by employing one-class classification (OCC) techniques, eliminating the need for non-rumour data during training. This method addresses challenges associated with inconsistent labeling of non-rumour data, aiming to enhance the reliability of rumour detection systems.

Objectives:

• Develop a rumour detection model using only rumour data for training.

• Mitigate issues arising from arbitrary labeling of non-rumour data in traditional binary classification methods.

• Evaluate the effectiveness of various one-class classifiers across different feature sets.

Detecting Rumours in Disasters: An Imbalanced Learning Approach

This project addresses the challenge of detecting rumours during disasters by applying imbalanced learning techniques. Given the scarcity of rumour data compared to non-rumour data in social networks, the project focuses on developing models that can effectively identify rumours despite this imbalance. 

Objectives:

• Collect and annotate datasets related to Hurricane Florence and the Kerala flood.

• Extract 83 theory-based and early-available features, including 47 novel features, to capture the properties of rumours and non-rumours.

• Develop rumour identification models using imbalanced learning techniques.

• Evaluate model performance through cross-incident learning, training on one incident and testing on another.

• Assess the impact of imbalanced learning using Bayesian Wilcoxon Signed-rank test.

Project: Bibliometric Analysis of Misinformation Studies

This project presents a comprehensive bibliometric analysis of misinformation studies, examining the academic landscape to understand its readiness and key research trends. By leveraging scientometric techniques and bibliometric data, the project maps the evolution, core topics, and collaboration patterns in the field.

Objectives:

• Conduct a large-scale bibliometric analysis of misinformation-related publications.

• Use scientometric frameworks to assess the academic readiness of the field.

• Identify key trends, influential works, and collaborative networks in misinformation studies.

• Develop visualizations and insights to guide future research directions.