I began Science Research looking into stock trends because ever since I was young, I enjoyed the stock market, especially watching my dad trade. I even had my own stock account with a few shares of Apple. From there, I have been researching predicting future stock trends using machine learning. I found a mentor and worked with him closely on using Valence Aware Dictionary and sEntiment Reasoner (VADER) analyzed Reddit posts to Predict the direction of Ethereum returns with a Long Short-Term Memory (LSTM) Network. I recently finished this project and have submitted my paper to the Regeneron Science Talent Search.

I proposed a means of predicting the future direction of Ethereum returns using the sentiment of Reddit posts with a Long Short Term Memory (LSTM) network. Reddit is a collection of communities where people can share information and opinions on various topics. It consists of many sub-communities called subreddits that discuss a particular topic. Ethereum is a decentralized blockchain platform that allows participants to transact with each other. We used Python Reddit API Wrapper (PRAW) to extract posts from a subreddit community called “ethtrader” for sentiment analysis because of its focus on Ethereum-related content. We used a Valence Aware Dictionary and sEntiment Reasoner (VADER), a Natural Language Processing (NLP) algorithm, to determine the sentiment of the posts due to its ability to recognize both the polarity and the intensity of emotions displayed in the text. We extracted market data using the Santiment API which provides metrics such as transaction volume, close price, and daily addresses. The sentiment and market data are the inputs in the LSTM. We trained the model on 64% of the data, validated it with the next 18% of the data, and tested it on the last 18% of the data. The model proved to be successful by predicting the direction of Ethereum returns correctly 75.47% of the time.

Currently, I am working to make my model faster and optimize the code for efficiency. Also, I am working on applying this model to stocks instead of cryptocurrencies to determine its success when applied to the stock market. I plan to attend college in 2023, double-majoring in finance and computer science and I want to continue research throughout college.