Our team members were interested in finance and machine learning and so we decided to explore the applications of ML in the stock market by developing a tool to make recommendations to help people invest more wisely. This tool can be a great resource that will streamline the investment process and make investment more accessible for newer investors as well as those who do not have the time or resources to dedicate to comprehensive research.
What it does
The tool recommends stocks to users based on their past history of investments and supports these decisions with calculations derived from the Goldman Sachs Marquee API's 4 scoring metrics which are indicators of a stock's projected growth, financial returns, valuation, and an integrated combination of the previous 3.
How we built it
Data Analysis: we implemented web-scraping where we parsed thousands of descriptions of stocks from Yahoo finance data and processed the data in order to consider the most significant words. Machine Learning: we used the Word2Vec machine learning algorithm which groups neighboring words within a block of text and computes the distance between words in order to determine the similarity between groups of texts. Marquee API: we used the API to access stock tickers as well as the 4 scoring metrics provided with the dataset, which provides users with forecast information for the stock.
Challenges we ran into
We spent the first 30ish hours experimenting (and failing) with numerous technologies, particularly relating to the cloud, we buckled down and basically completed the entire project in the final 8 hours or so. Furthermore, we are a very new team as this was the first full hackathon experience for 3 of 4 members.
Accomplishments that we're proud of
Despite a tumultuous day of trial and error, we finally have a working product that incorporates machine learning and the Marquee API, thus satisfying our initial goal of exploring the applications of ML in finance
What we learned
In addition to the wealth of technical skills acquired from our explorations with web development and machine learning, we also got a taste of the hackathon experience and the stress involved with making a project from end-to-end in a fast-paced environment.