Meta AI

Machine Learning stock prediction modeling website using a polynomial regression, allowing investors to easily predict small changes in the market

Comment

Inspiration

I started noticing that there were a lot of cyclic patterns in stock charts and I did some research and learned that data scientists have used ML regression models and the patterns found in stock charts to predict future stock prices for a couple of years now, and I wanted to create my own models to test if it could be profitable

What it does

Essentially, Meta AI is a website with a database of some of the most trending stock right now. Using python and regression models that I've tuned, the website displays the most recent and accurate polynomial predictions of stock prices for these companies for the next few days

How I built it

There were 3 major parts in building this hack. First, I had to build the actual regression model which was probably the hardest part. I had to research the different types of regression models, choose the one that worked best, and coded the arrays and CSV files that I had of stock data to be plotted on the model. Next, I took the predicted values and placed them in firebase in order to move it to my javascript-based website later. Since I'm well versed in HTML and CSS, the last part of creating the actual website was quite simple. It was just connected to the firebase realtime database for the text and pictures, and I made a quick aesthetic design

Challenges I ran into

The largest challenge I ran into was initially getting the regression model to work. I tried 3 different regression models: linear, exponential, and decision tree before I finally tuned and settled on polynomial regressions and decided it was the best one. This alone took me a large chunk of the entire work time: maybe 5-6 hours.

Accomplishments that I'm proud of

I'm proud of being able to create the entire design of the website in 3-4 hours and integrating it with the firebase data very efficiently.

What I learned

Most of the regression techniques, tuning, and packages that I used were all completely new to me. I have plotted arrays and created equations before, but ML regressions were completely new to me so I had to learn as I went along with the project.

What's next for Meta AI

There are a lot of improvements I'd like to make to Meta AI after this hackathon, and after I get some sleep. First of all, I'd obviously like to expand. Currently, there are only 15 stocks being displayed, and Id like to make that at least 50. Id also like to add a search and in-built API function where visitors can search up the model for a specific stock and it calculates the regression on the spot. Id also definitely like to include SENTIMENT ANALYSIS. As I mentioned in my video and on my website, mathematics and financial data is only one factor in predicting stocks... using sentiment analysis, we can gauge a better understanding of what society thinks of the company which can give us a better understanding of where it's going as well.

Share this project:

Updates