Eric Vistnes

Face Mask Detection and Recogntion (2021), an Image Processing model based on VGGNet, a Convolutional Neural Network (CNN), trained to classify faces based on the presence of face masks. I used modified Haar Cascade logic to find faces in real world images and achieved a validation accuracy of 99.15%. Implemented in a program, it locates and classifies real-world faces in a usable interface.

Heart Disease Predication (2021), a supervised Classification study trained on a global dataset of patients with heart disease. Models perform with accuracy of 87% and sensitivity of 94%, matching and exceeding leading models. The study highlights the need for more diverse research from varying medical centers.

Plantr (2021), created a prototype for a Human Computer Interaction project to design, review, and create a plant care app. Tested with various review groups and rapidly iterated with new feedback. Coded in Axure and Figma.

FIFA Player Analysis (2021), my team and I used Principal Component Analysis to categorize all player statistics into the fice components of Attacking, Defending, Hype & Value, Players Physique, and Others. Exploratory Factor Analysis revealed that player reputation and compensation are not linearly related. Finally, we implemented Canonical Correspondence Analysis confirmed that player's skills and attributes are strongly correlated to their position.

Netflix Prize Data (2020), my team created a recommender system from the Netflix Prize data of 100,480,507 movie ratings. I reduced data usage by 80% through data cleaning, ranking, and pickling. We used K-Means Clustering and K-Nearest Neighbors to recommend similar movies to similar users with reinforcement learning. We used Singular Value Decomposition to reduce dimensionality as well as to predict specific movie ratings by individual users.