Stock Trend Prediction

Stock prediction accuracy (5-fold CV) for Apple

Inspiration

We are both interested in investing in stocks but this is our first time working with financial data. We decided to start with the basics by seeing if we could use the given Goldman Sachs dataset to predict trends in the stock market.

What it does

The user can select a company (we selected Apple in our submission) and define a time window for which he/she would like to know the stock market trend. The algorithm will predict whether stock prices will increase or decrease with 60-94% accuracy. The accuracy depends on the time window selected by the user.

How we built it

We used the scores provided in the Goldman Sachs dataset along with the information we pulled from IEX to use as features. We then used the selected features to do supervised learning using binary labels (+1 for increasing and -1 for decreasing stock closing prices). We used SVM with a Gaussian kernel as our classifier as it performed best.

Challenges we ran into

This was our first time working with financial data so getting used to the terminology took some time.

Due to the storm, we were unable to get started with our hacking until Saturday evening as we were trapped inside a train for over 12 hours with spotty wifi, and at 3 am one of our teammate's computer crashed, which severely hindered our progress.

Accomplishments that we're proud of

We're able to predict with fairly high accuracy 70-90% depending on the time window. Considering our circumstances we are proud that we were able to build a working model. We also built an interface where the user can select a company from the drop down menu and select a time window to determine how accurate the stock trend prediction will be.