Data Science | Process Analysis | Business Intelligence Will Griffin

Recommender System "Nextale" Capstone Project

In my final capstone project with General Assembly's Data Science Immersive program, I built a recommender system to provide Book, Movie, and Video Game recommendations for users based on their search term(s). I also added an exclusion term functionality so that users could add a keyword to filter out of any search results, allowing for more customization and ultimately results that more accurately reflected the user's goals. After building the recommender, I was able to deploy it online using streamlit.io.

Try it out for yourself!

Link to github repo

Natural Language Processing, Logistic Regression, Random Forest, Naive Bayes Reddit Webscraping Project

Photo by Kamil S on Unsplash.com

In this project I scraped 20k Reddit posts from pushshift.io - 50% each from the Xbox and Playstation subreddits. I performed cleaning, EDA, and vectorization to build Naive Bayes, Random Forest, and Logistic Regression models, classifying posts between the subreddits with 92% accuracy. I also identified the most frequent words in each subreddit to explore differences in the conversations had by each player base and make marketing recommendations.

Link to github repo

Linear Regression, Time Series The Effect of News Sentiment on Stock Performance

Photo by Adeolu Eletu on Unsplash

In March 2021, some former DSI classmates and I entered a Data Science competition through the Global Association for Research Methods and Data Science "to analyze the effects of news sentiment on daily stock performance for top companies in the oil and gas industry." It was an opportunity to work with a wealth of real data on time-series modeling and stock prices, two elements that none of us had much prior experience with. We compressed and formatted our work into a fun, ineractive report using Streamlit.io - check it out for yourself!

Web App

Link to github repo