2023-12-06: MSV Student Capstone Conference (MSVSCC 2023) Trip Report

By -

This year, the Modeling, Simulation and Visualization Student Capstone Conference (MSVSCC 2023) celebrated its 16th anniversary. The MSV Student Capstone Conference showcases the practical research work of both undergraduate and graduate students in modeling and simulation, encompassing various disciplines from different departments within Old Dominion University (ODU), as well as programs from other colleges and universities. Students were required to submit research papers (either a two page extended abstract or a 12-page full paper) from any of the seven conference paper tracks: Transportation, Business & Industry, General Sciences & Engineering, Education & Training, Virtual Environments & Visualization, Infrastructure Security & Military, Medical Simulation, and Data Science. Papers were reviewed by research and faculty track chairs from ODU and the Virginia Modeling, Analysis, and Simulation Center (VMASC).


The conference was held as an in-person conference on April 20, 2023 at VMASC. It was our first time attending the MSVSCC and representing the WS-DL research group at the conference to present our work. The conference had morning and afternoon sessions. The Data Science track was introduced this year (2023). Here, we have summarized the presentations of the Data Science track that were covered in both sessions. 

Papers

  1. Assessing the Frequency and Severity of Malware Attacks: An Exploratory Analysis of the Advisen Cyber

    Loss Dataset 

Authors: Ahmed M. Abdelmagid, Farshid Javadnejad, C. Ariel Pinto, Michael K. McShane, Rafael Diaz, Elijah Gartell

  1. The Effectiveness of Visualization Techniques for Supporting Decision-Making 

Authors: Cansu Yalim, Holly A. H. Handley

  1. Extracting Information from Twitter Screenshots

Authors: Tarannum Zaki, Michael L. Nelson, Michele C. Weigle

  1. Behind Derogatory Migrants’ Terms for Venezuelan Migrants: Xenophobia and Sexism Identification with 

    Twitter Data and NLP 

Authors: Joseph Martínez, Melissa Miller-Felton, Jose Padilla, Erika Frydenlund

  1. The Legacy of Colonization and Civil Societies in South Africa

Authors: Erika Frydenlund, Melissa Miller-Felton, Bolu Ayankojo

  1. Assessing Frustration Towards Venezuelan Migrants in Columbia: Path Analysis on Newspaper Coded Data

Authors: Brian Llinas, Guljannat Huseynli, Erika Frydenlund, Katherine Palacio, Jose Padilla

  1. Exploring Xenophobic Events through GDELT Data Analysis

Authors: Himarsha Jayanetti, Erika Frydenlund, Michele C. Weigle

  1. GPU Utilization: Predictive SARIMAX Time Series Analysis

Author: Dorothy Dorie Parry

Presentations

Presentation 1

Ahmed M. Abdelmagid and Farshid Javadnejad from the ODU Department of Engineering Management and Systems Engineering presented an extended abstract paper titled “Assessing the Frequency and Severity of Malware Attacks: An Exploratory Analysis of the Advisen Cyber Loss Dataset”. The objective of their project is to employ diverse data science techniques for assessing the vulnerabilities associated with potential cyberattacks on the shipbuilding supply chain.


To achieve this, the authors made use of the Advisen Cyber Loss Data to demonstrate the feasibility of categorizing risk patterns within the shipbuilding supply network. Their results indicate that cyber attackers focus their efforts on organizations of various sizes, encompassing both large and small entities. This paper gives useful insights into the changing cyber threats, emphasizing the need for organizations to be watchful in their cybersecurity efforts.



Presentation 2 

Cansu Yalim from the ODU Department of Engineering Management and Systems Engineering presented a full paper titled “The Effectiveness of Visualization Techniques for Supporting Decision-Making”. The authors focus on investigating the effectiveness of various visualization techniques for different data types, such as continuous, categorical, and time-series data. In the paper, they discussed different plots (common plots such as line charts, bar charts, scatter plots, box plots, and histograms, alongside less common plots like violin plots, kernel density estimation plots, and wavelet analysis charts) and for each type, the authors examine their strengths and weaknesses. The main objective of the paper is to address the absence of precise recommendations for choosing the most suitable data visualization method for distinct data types.


Presentation 3 

Tarannum Zaki from the ODU WS-DL research group presented an extended abstract paper titled “Extracting Information from Twitter Screenshots”. The prevalence of screenshots of fabricated information on social media for information sharing raises concerns about the potential spread of misinformation and disinformation. Users often share screenshots without verifying their authenticity, leading to the spread of fake content. To address this issue, the authors are developing a tool to automatically assess the probability of a screenshot being misattributed by leveraging services from the live web and web archives. Providing users with a misattribution probability assessment may encourage them to adopt greater caution before sharing screenshots and help prevent the inadvertent spread of misinformation. The paper discusses methods for extracting metadata from Twitter screenshots (e.g., timestamp) in order to prepare the input data for the tool to be developed. 


The presentation for this paper won the Best Presentation award in the Data Science Track.



Presentation 4 

Joseph Martínez from the ODU Department of Computational Modeling & Simulation Engineering and Melissa Miller-Felton from the GPIS at ODU College of Arts & Letters International Studies presented their work titled “Behind Derogatory Migrants’ Terms for Venezuelan Migrants: Xenophobia and Sexism Identification with Twitter Data and NLP”. The authors explored how the entry of over 2.5 million Venezuelan migrants into Colombia resulted in tensions, creating new challenges and negative sentiments within communities. In their work, they analyzed 5.7 million tweets from Colombian users spanning from 2015 to 2021 by tracking derogatory terms like "veneco" and "veneca" used for Venezuelans. They discovered that the annual tone distributions for both terms mainly displayed a negative tone, with the highest peak occurring in 2018, which corresponded to a significant influx of Venezuelan migrants into Colombia. The definition of "veneco" has evolved over time, whereas "veneca" has consistently carried a negative meaning. Negative tone signifies intense emotions, not necessarily an adverse perspective towards migrants. Events like soccer matches, political incidents, and statements from politicians affected the negative associations with the term "veneco." Meanwhile, the negative connotations tied to "veneca" depicted Venezuelan migrant women in a negative way, worsening damaging stereotypes, which in turn made them more excluded and subjected to social stigma.



Presentation 5 

Melissa Miller-Felton and Bolu Ayankojo both from the GPIS at ODU College of Arts & Letters International Studies presented their work titled “The Legacy of Colonization and Civil Societies in South Africa”. The authors addressed how the influx of migrants into Khayelitsha, a township located in the Western Cape of South Africa, has resulted in strained resources. They used computer models to understand how migrants and local people interact, based on interviews with 24 people in Cape Town from 2020 to 2021. For their preliminary work, the authors used a tool called Loopy, which allowed them to simulate the behavior of real-world situations. More precisely, they established nodes representing actors such as NGOs, Government, Migrants, and Locals, and examined how these entities interacted with each other. Their findings indicated that a reduction in local resource requirements corresponds to a rise in xenophobic sentiments. A decrease in the number of NGOs results in reduced information availability, and diminished government involvement leads to fewer resources and heightened xenophobia. Consistent NGO participation within the community, on the other hand, would lead to increased access to affordable housing, job prospects, and educational opportunities. The authors wrapped up their presentation by noting that their work is a preliminary model for an Agent Based Model (ABM).


 


Presentation 6

Brian Llinas from the ODU WS-DL research group and Guljannat Huseynli from VMASC presented their work titled “Assessing Frustration Towards Venezuelan Migrants in Columbia: Path Analysis on Newspaper Coded Data”. Their study analyzed 1,360 articles extracted from local and regional newspapers in Colombia between 2015 and 2020 to examine the influence of Venezuelan migrants on levels of frustration in locals in Colombia. These articles were previously qualitatively coded and categorized into frustration types such as Frustration towards Migrants, Infrastructure, Government, and Geopolitics. Employing path modeling for statistical analysis, they revealed that an increase in migrant arrivals was positively correlated with heightened frustrations among locals towards migrants, the Colombian government, and infrastructure. However, an increase in migrant arrivals was associated with reduced frustration towards geopolitics among locals. The authors pointed out how these findings hold significance for future implications, including the need to develop effective policies and programs, enhance investment in public services, foster international cooperation, and launch education and awareness campaigns. 


 

Presentation 7

Himarsha Jayanetti from the ODU WS-DL research group presented their work titled “Exploring Xenophobic Events through GDELT Data Analysis”, an exploratory study where the authors looked into xenophobic events related to refugees and migration using the Global Data on Events, Location, and Tone (GDELT) news database. They conducted this analysis using visualizations guided by  two case studies. The initial case study revolves around the surge in news coverage concerning refugees following the incident of the death of Alan Kurdi, a two-year-old Syrian boy. The sentiments of news articles were consistently negative both prior to and after the incident, with a notable increase in the range of the average tone (maximum – minimum) post-January 2015. As for the second case study, the authors investigated a sudden increase in news coverage during March 2021, initially assuming it was related to the Atlanta spa shooting. Nevertheless, their visualizations (a choropleth map and bar chart) identified Spain as the main country, with the surge being attributed to the increased influx of African migrants to the Canary Islands. Although the authors outlined the challenges and limitations of the analysis, they emphasized the significance of investigating and understanding xenophobic events as a pivotal step in addressing the global issue of xenophobia. 


This paper won two awards at the conference: Best Paper in the Data Science Track, and Best Overall Paper.


 

Presentation 8

Dorothy Dorie Parry from the ODU Department of Electrical & Computer Engineering presented a full paper titled “GPU Utilization: Predictive SARIMAX Time Series Analysis”. The current demand for enhanced performance, data processing, and computational capabilities corresponds to an increased need for power and energy consumption. The goal of this work is to understand how Inception3 (a memory-intensive parallel image classification algorithm) behaves in terms of resource usage and performance over time and to develop predictive models based on these insights. The performance data (such as memory, GPU utilization, power consumption, and GPU temperature) for the experiments were gathered using a profiler tool called "nvidia-smi". Parry applied a time series model called Seasonal Autoregressive Integrated Moving Average Exogenous (SARIMAX) to predict and understand patterns in GPU utilization over time. Parry noted that the accuracy of the predictions fell short of their expectations, citing factors including irregular fluctuations in the time series patterns. Nonetheless, this analysis has established a foundation for future work aimed at improving resource allocation that involves investigating enhanced techniques for time series data prediction, utilizing Transformer Neural Networks with larger historical datasets for training, enabling real-time predictions, and utilizing predicted GPU utilization data to enhance resource and power allocation.



Awards

As the day came to a close, the much-awaited conference awards were announced. There were awards for both Best Presentation and Best Paper in each individual track, along with recognition for Best Overall Paper across all seven conference tracks. Tarannum Zaki won the award for the Best Presentation in Data Science Track for “Extracting Information from Twitter Screenshots”. Himarsha Jayanetti won Best Paper in Data Science Track as well as Best Overall Paper awards for “Exploring Xenophobic Events through GDELT Data Analysis”. The conference wrapped up on a high note, spotlighting the exceptional research and innovation by student authors, while also serving as a valuable platform for them to share their work with the modeling and simulations community.

 


Conclusion

This was the first time Data Science track was introduced in the MSVSCC 2023 conference and the first time for WS-DL members to participate, too. It was a great experience to share thoughts and gain insights about the diverse research that is going on relevant to Data Science at ODU. We would like to express our gratitude to the organizers of this capstone student conference for providing such a wonderful opportunity to share our research work. We are extremely happy and humbled to have won all the awards from the conference for the Data Science track. We are looking forward to presenting at the conference next year!


 –  Himarsha Jayanetti (@HimarshaJ) and Tarannum Zaki (@tarannum_zaki)

Comments

Post a Comment