My project, EquiTrade, uses machine learning to predict stock prices on the S&P 500 using historical time-series data and a deep learning regression model. It uses data on stock closing prices, various financial indicators and more to make predictions on future prices. I used Tensorflow and Google Colaboratory to develop the base product, but as this project begins to grow more and more computationally tasking, I am looking to expand to work with a lab to train my model. The model accepts a 2-dimensional matrix as input, which consists of the sample of data and the number of stocks for which to predict for, which is 500. The model then implements a series of weights and biases to the data to recognize and avoid any bias and noise. Finally, the model outputs the stock prediction which is compared to the correct price and fed back into the model to improve its accuracy, using an optimizer. The reason why I developed this application is because I saw the importance of quantitative trading, as over 70% of trades are generated by computerized platforms. Not only this, but I believed in the key principle that you are your own greatest investment manager, and that machine learning prioritizes one thing: to generate returns. Instead of trusting our money in the hands of firms that have other interests, such as pleasing creditors, stakeholders, and managers, I believe that ML will allow for consumers to not rely on these companies and systems for wealth generation.
What inspired you (or your team)?
I was reading the Financial Post one day, and happened to see an article about quantitative trading and Two Sigma, a hedge fund managing over 50 billion in assets using purely AI to do so. I was hooked. I spent the past 1-2 months reading on algorithms, models, network architectures, backpropagation and more and began developing EquiTrade as a solution I hoped I could release as direct-to-consumer. I realized that machine learning has the power to revolutionize finance and the way we work, and only in the past few years are banks and other firms beginning to take notice and build algorithmic trading products. I wanted to ideally create something that could not hold bias like the human brain does, and allow for any consumer regardless of previous financial knowledge to understand. In the future, I am working towards building an online cloud platform to release it on, and use Natural Language Processing (NLP) models to read financial data and reports to aid in investment decisions. In my product, I imagine a future where consumers are more in control of their finances, instead of relying on others to do so, and my goal is to create a product that produces at least 5-10% in returns within the near future.