ML/AI - S&P 500 Price Predictor

We used machine learning and the last 9 months of Covid-19 and stock market data to predict the S&P 500.

Comment

Inspiration

We were very interested in looking at different factors and how they affect one another. While searching for COVID numbers we realized there was a spike in new deaths around the same time there was a large drop in the S&P 500 price. Wanting to use machine learning this was a perfect challenge for us.

What it does

We used a custom machine-learning algorithm to create a model for predicting the price of the S&P 500 for the following day. Our script uses the current day's COVID stats to predict the following day's S&P 500 numbers. Then it goes on to predict the numbers for the past 9 months and outputs some nice graphs comparing the true values to the predicted ones.

How we built it

We built it using the industry-standard Tensorflow 2.0 by google along with Python and React.js. These together make up the machine learning algorithm in Tensorflow with Python and the website which uses React in JavaScript.

Challenges we ran into

Some challenges we ran into include finding data that had any correlation or are good for Machine learning, communication with the front end and making graphs, and tuning the machine learning algorithm to find the best results. These challenges forced us to learn valuable data science skills and apply them to real-world examples.

Accomplishments that I'm proud of

We are proud of our algorithm that predicts the values, the graphs that display the values, and the algorithm that makes the model. We put a lot of time into all of these components and many others and are very happy and proud of the outcome.

What I learned

We learned a lot from this project. One thing that we learned was how to use a machine learning model as the back end for a website with node.js front end. We also got experience with the various file formats for the machine learning libraries, such as .h5 to store TensorFlow models and .save for saving scalers. Finally, we learned that even though Google Colab lets you use their higher-powered machines so learning is faster in theory, in many cases such as ours, running it locally in a Jupyter notebook can actually be much quicker.

What's next for ML S&P Price Predictor

The next step we want to take for ML S&P Price Predictor would be to use it for a stock trading bot based on the data for US COVID cases. We would enter the stats on COVID from the day, have the machine learning model predict the closing price of the S&P 500, and then make a one day bet on the fund based on that.

Built With

Share this project:

Updates