WolframWallstreet

A cutting-edge ML Training and Financial Simulation tool tailored for stock market trading, with a special focus on accurately predicting the performance of automotive industry stocks.

Comment

Inspiration

Our motivation originated from wanting to create a tool that is accessible to anyone in the community who wants to profit in the stock market without an extensive level of financial expertise. Financial platforms are generally not accessible due to their complexity and exclusivity, but using our tool, any user can make an informed investment decision.

What it does

Name a stock! Our tool will utilize historical data for that stock up to 2015 to train a machine learning model. After the model is complete, it will simulate the buying and selling of the stock from 2015 to 2020 and conduct data analysis to determine the profitability of the model's trading strategy and how it compares to the performance of a simple buy-and-hold strategy.

How we built it

Our tool's front-end is developed using JavaScript with React.js, while an Express.js server drives this website. When a request is made, the Express server initiates a Python script to power the ML models. We tried several different types of ML, including Linear Regression, Random Forest Regression, Gradiant Boosting Regression, and Neural Networks. In the end, the Linear Regression model performed the best, so we chose it for our simulations.

Challenges we ran into

At one point, we were making extremely high returns on every stock (sometimes up to hundreds of thousands of dollars from a $10,000 initial investment). While we were happy with our results, we knew that something wasn't right. If it was that easy to beat the stock market, everyone would be doing it already. Eventually, we figured out that one of the parameters in our ML model was getting information about the future that it wasn't supposed to have. After fixing that issue, our results returned to the realm of reality.

Accomplishments that we're proud of

While our model does not do well in high growth stocks (the technology sector, for example), it is extremely good at trading automotive stocks. The more stable nature of automotive stocks means that a buy and hold strategy does not net high returns, lowering the bar for our model to beat.

What we learned

We learned a lot about the stock market and the mechanisms that make it work. This knowledge will help us in our future with planning for retirement and being financially literate. We also learned about the many different ML training techniques that we tried for this project. There is no one right way to train an ML model, you have to try different methods and see what works.

What's next for WolframWallstreet

Now that we have a training method that works for automotive stocks, the next step is to generalize our training methodology to work in the broader market. We can also work to make this technology more open and accessible so that other people can learn about the financial systems that make our economy tick.

Share this project:

Updates