Hi, I am Dean Roeser

Twitter Stock Predictor

Python | Machine Learning

This project was split into three distinct parts: parsing tweets, building the sentiment analysis, and using the sentiment analysis to predict price. The language processing is primarily handled by cleaning the tweets, and using the NLTK package to format the tweet into a usable bag of words. Similarly, predicting stock prices is a matter of statistically analyzing historical stock prices gathered using the Yahoo Finance API and information from our sentiment analysis. The primary AI algorithm we used in this project is the sentiment analysis. Sentiment analysis works as a proxy to understanding the content of a sentence. It imitates human language processing by analyzing the pieces of a sentence, such as words and punctuation to determine the general negative or positive connotation of the sentence.

Check out the code for this project here!

Global Game Jam 2023

C# | Unity

Putting my new Unity and C# skills to the test, I joined a random team for Global Game Jam 2023. The theme was "Root", so we designed and programmed a tower defense game where you play as a dig-dug-esque character and defend your crops from worms, ants, and moles with a variety of traps. Creating an entire game in one weekend while dealing with Unity gitignore problems, merge issues, and creative differences was a fun challenge. I can't wait to do it again next year.

Check out the code for this project here!

NBA Predictive Modeling

Python | Machine Learning

Predicting the winner of an NBA postseason is challenging. However, we were able to generate a decision tree model using SKLearn. We trained our model on 10 years of regular and post-season data, and tested it on the current year, going back to 2001. Our model accurately predicted the championship game from 2012, 2015-2018, and 2021 based solely on regular season data. When we completed this project in 2022 with only half of the season's data, the two teams in the finals were in our predicted top 4.

Check out the code for this project here!

Autonomous Robotics

C++ | Vex Robotics

Given a month to build and program an autonomous robot to solve a blind maze faster than the 11 other groups, our team quickly built a quick, nimble, two-wheeled "scoot-bot" with the limited materials we had. Using a combination of active sensing, PID control, and experimental mapping, our bot was the fastest and most consistent of all. 

Unity Game Modification

C# | Unity

While many school assignments begin with creating a new file, this series of assignments for my C453 class involved finding "completed" games on game jam websites like itch.io and modifying them. Getting familiar with other programmers' code to find bugs and add new features is something not often taught in schools, but is a vital skill to succeed as a programmer.

Check out the code for these projects here!