EquitoPredicto

Inspiration

Being stock market and finance enthusiasts, we wanted to build a project that would allow us to experiment with tools related to this fields.

What it does

EquitoPredicto attempts to predict trends in a company's stock prices based on historical patterns.

How we built it

Our algorithm works by mapping the pattern created by the last 200 days of the stock's price on the last 20 years of the company's history to find similarities. This works by calculating the linear regression of our 200 day pattern in order to represent it as a function. Once our pattern is represented as a function, we compare to multiple different sets of 200 days in the past 20 years and select the best matches by using their mean squared error and mean average error. Then, our prediction is calculated from the mean of all the selected matches and the max and min curves are the max and min values for every day of our prediction.

Challenges we ran into

A challenge we encountered is to analyze data from the past years of a company while keeping our algorithm fast enough. Furthermore, the time constraints limited our capacity to test and improve our algorithm as much as we would have wanted.

Accomplishments that we're proud of

Despite the short amount of time and this being our first project of this type, our algorithm manages to predict somewhat accurately trends in stock prices.

What we learned

We learned how to create a project using a frontend in HTML, CSS and JavaScript, and a backend using Python. Furthermore, this project allowed us to understand better concepts of trends and patterns in trading.

What's next for EquitoPredicto

In order to improve our project's accuracy, we would try to use many patterns of different size to analyze historical data. Also, we would try to implement a statistical approach to attempt to balance our predictions.