Title of presentation: Pattern Analysis of Flight Delays  
Presenter:  William Robert Bartels
Advisor:  Dr. Karen Works
 
Abstract:

Are there any regular patterns to flight delays?
In this exhibition study, I utilized a data set of reported flight delays and created a graphical database of such events using NEO4j. The Flight Delays dataset was obtained from Kaggle (https://www.kaggle.com/datasets/flight-delay). It is comprised of flight details, delay information, and weather data. As preliminary analysis I examined distributions and trends using visualizations in Python. I then created a graphical database modeling airports as nodes and flights as relationships to analyze potential delay propagation patterns . I am exploring how to structure the graphical database to search for common patterns in scheduling sequences that lead to flight delays.

VIDEO

Title of presentation: 'NextGenProperties' : A Data-Driven Approach to Automated Real Estate Valuation and Categorization  
Presenter:  Renzo Broggi
Advisor:  Dr. Karen Works
 
Abstract:

Current real estate valuation tools stand to benefit from additional insights beyond those based solely on analyses of sales data. This project, NextGenProperties, seeks to address the need by introducing an automated tool for analyzing real estate sales data and categorizing properties based on various factors such as location, size, market trends, and investment potential. By leveraging machine learning algorithms, our system processes large datasets to extract meaningful insights, allowing users to make informed decisions regarding property valuation and investment opportunities. By implementing a combination of data preprocessing techniques and feature extraction methods, 'NextGenProperties' seeks to streamline the analysis process and provide accurate estimations, while minimizing manual effort. As an exploratory submission, my presentation will showcase the potential 'NextGenProperties' has to help revolutionize real estate estimation through data-driven automation. 

VIDEO

Title of presentation: Effects of Graph Augmentation of Graph Neural Networks (GNN) Twitter Bot Detection Models  
Presenter:  Juan Sanchez Moreno
Advisor:  Dr. Karen Works
 
Abstract:

There is a growing concern about the presence and real proportion of bots in social media, especially on Twitter (now X). These bots can spread misinformation, impose narratives, and distort the reality of the platforms users. Graph Neural Networks (GNN) Twitter bot detection models have been shown to be highly effective. We seek to improve the detection accuracy of social media bots and the efficiency of the GNN by adapting the relationships between nodes and randomly dropping nodes and edges on the graph.  Our experiments bound the levels at which each of the aforementioned adjustments stop improving the GNN.

VIDEO