Student Presentations: INFORMS Annual Meeting 2020

Rail Accident Data Analysis: Coldspot Analysis

Christie Nelson, John F. Betak, Trefor Williams, Prajwal Chadaga Krishnamurphy, 

Hotspot and cold spot analysis was performed for 4 regional railroads to understand the attributes that may or may not factor in accidents. This data was mapped in GIS and analyzed with datasets from FRA, Google Earth, and more.

GIS and Visual Exploration of a Class 1 Railroad & Accident Data

Arshiya Verma, Christie Nelson, Devam Shah, Namrita Gupta, Akanksha Sharma, Cyrus Yu, John F. Betak, Trefor Williams

We obtained data for a Class I railroad in the United States to perform visual and GIS analysis of its accident data. We created a framework of how to obtain various attributes from FRA Accidents data, Equipment, Inventory, Google Earth, and more. From there, we identified and mapped hotspots of accidents, and identified underlying correlating factors associated at each specific grade crossing location.

Digital Forensics Career Advancement Requirements and Job Pathways

Ryan Aponte, Christie Nelson, Dennis Egan, Fred Roberts

Focusing on digital forensics for FLETC and DHS, and building on prior work, we researched job pathways by level and rank, what skills/certifications are needed by rank, and what is needed to advance. We have also gathered employment text data, and performed LDA and text modeling, to gain additional insights into our career pathways. We supplement this work by adding what skills, courses, and certifications are required as an employee moves from entry-level to mid to advanced.

Analyzing COVID-19 Press Conference Transcripts Through Topic Modeling and Sentiment Analysis

Therese Azvedo, Christie Nelson

Across the nation, state governors have held press conferences related to the pandemic. Topic modeling, word frequencies, and sentiment analysis were conducted on press conference transcripts from various states. By incorporating these techniques, weekly themes and word frequencies were revealed as well as monthly analyses of sentiment on the press conference transcripts.

Geographic Clustering to Identify Similar MO for Homicides with Female Victims

Hedy Makris, Christie Nelson

This study proposes a method for identifying clusters of female homicides with similar MO-related attributes in close geographic proximity and recommending them to law enforcement for further investigation. A dataset of homicide incidents were converted into a numerical array of features describing the victim’s physical characteristics, murder weapon used, date, location, and any circumstances surrounding the homicide. A maximum dissimilarity threshold calculated based on the pairwise Euclidean distances between each case was used to establish clusters of potentially linked female homicides.