Student Capstone Presentation

12:00 - 12:50 PM, Wednesday, April 29, via ZOOM
(contact Dr. Kerby for Zoom link)

Implementing SQL for Table Toolz, Project Structures for Data Science Daniel Lew
I will be talking about implementing SQL for Tabletoolz, which is a data manipulation library inspired by R’s Dplyr in Python, what I did and what I learned from the project. Also, I will be sharing about using a project structure for data science.

Internship Presentations

12:00 - 12:50 PM, Friday, April 24, via ZOOM
(contact Dr. Kerby for Zoom link)

Internship Experience: Supply Chain Intern at Fastenal Danielle Buoy
Fastenal, an industrial supplies company based in Winona, MN has worked for years distributing goods used by hundreds of businesses. Inventory movement is emphasized in their supply chain department, where I’ve contributed my internship skills. Handling product redistribution amongst the branches, facilitating vendor returns, and assisting with liquidations of non-moving parts are three main areas of focus. An essential part of the process as a whole is to minimize and reduce noncontributing inventory via utilization of programs, tools, and databases. Optimization of productivity and profitability throughout the company is a major takeaway of my supply chain duties.
Internship: Fastenal Supply Chain and Data Team Peyton Schumacher
	I will speak about my duties at Fastenal as a Supply Chain Analyst and Data Team Member, as well as creating and refining new and existing processes in our Supply Chain.

Student Capstone Presentations

12:00 - 12:50 PM, Wednesday, April 22, via ZOOM
(contact Dr. Kerby for Zoom link)

Methods of Data Imputation Carson Howells
Imputation is the process of filling in missing values in data sets. This project is exploring which method of data imputation is best for different data types.
Deep Learning Classification of Retina Images: Detecting Diabetic Retinopathy Mikolaj Wieczorek
	Diabetic retinopathy (DR) is a common eye disease that causes vision loss for people with diabetes. It is one of the fastest growing causes of preventable blindness, and about 45% of Americans diagnosed with diabetes have some stage of DR. As the disease has no symptoms in early stages and treatments begin after the vision is already somewhat affected, early detection is of pivotal value to diabetic patients. The aim of this project is to build a deep learning classification model that could assist in early detection and severity grading of DR. The final results are obtained by implementing a Dense Convolutional Network Architecture model that yields 98% classification accuracy. Further study is being considered with application of larger training data and to deploy the final model as a web app.

Student Capstone Presentations

12:00 - 12:50 PM, Wednesday, April 15, via ZOOM
(contact Dr. Kerby for Zoom link)

Using Natural Language Processing to Analyze News Bias Bradley Erickson
As we live in the information age, finding reliable information becomes more and more difficult. All news sources contain bias, thus anything someone reads could be misinformation or interpreted differently. This research scrapes thousands of articles from many sources relating to 5 controversial topics: the impeachment, abortion, immigration, gun control, and climate change. We use unsupervised learning techniques to rank and filter out the top sentences by giving more weight to unbiased, more credible sources. This information will form an extractive-based summary. Finally, we compare the language used among 3 bias groups: left bias, no bias, and right bias. The research provides insights into the current bias state of news reporting.
Social Influence and the Freshman Experience Adam Funk
	We investigated the relationship between time at a university and who freshmen are most influenced by at that given time. The study took a subset of 1000 freshmen from the Winona State University class of 2023 and asked about different influence factors and changes in behavior over the course of 10 weeks. Different surveys were sent out at 1 week, 3 weeks, 6 weeks, and 10 weeks. Questions on who they were influenced by, both negatively and positively, were asked, as well as groups they were a part of and whether they felt they had changed as a person. We also attempted to monitor the change over time of specific individuals through an identification code, which was unsuccessful. However, summary statistics of each time point were successfully collected and reported.

Student Capstone Presentations

12:00 - 12:50 PM, Wednesday, April 8, via ZOOM
(contact Dr. Kerby for Zoom link)

Identifying Pulsar Stars Alexander Phillips
Pulsar stars, neutron stars that appear as rhythmic pulses of radio waves, are of great importance to astronomers as measuring devices. Scientists from two observatories in Australia and Germany compiled a dataset of 17,898 potential pulsar signals, recording data on each signal’s integrated profile and dispersion measure. This study considers variations on a logistic regression model to classify radio signals as pulsar stars or noise. The optimal predictive model considers the mean, standard deviation, skewness, and kurtosis of both the integrated profile and the dispersion measure, though a model with only the four best predictors also reaches a high level of accuracy.
Analyzing Filings Sentiment for Applications in Finance Marshall Will
	There has been increasing use of using sentiment in financial reports to create a better understanding of what is being read and how that can affect market prices. Analyzing sentiment for financial research has been around for a while, but over the past decades, there has been an increasing interest in looking at how changes in sentiment can affect market prices. Past research has shown that measuring sentiment can be used to generate Alpha and gauge a firm’s fundamentals. For my capstone, I looked at the sentiment of 10-K and 10-Q filings for publicly traded securities in the NASDAQ and NYSE from 1993 to 2018 to see if it is possible to predict if a security will rise in price based on the change in word sentiment in these public filings.