Future Stock Price Prediction

Inspiration

When coming up with ideas for this project, we were interested in the idea of time series forecasting and felt stock prediction fit into this idea very well

What it does

Given an input of stock prices, this application will attempt to continue building data into the future

How we built it

Using a machine learning library I built, data scraping techniques to get historical prices from the Nasdaq website, and Java JFrame applications, we were able to build a desktop application capable of training and testing on given datasets, all with just a few presses of a button.

Challenges we ran into

We had numerous problems throughout the way, such as finding how to make the data scraping algorithm work properly and many many problems associated with JFrames. Throughout the construction of the ML library, there were numerous situations in which problems that seemed so basic would take increasingly longer times to solve. Surprisingly, however, there were several situations that went far more smoothly than expected. The windows created were surprisingly cooperative, and many of them did not take long to make fully functional.

Accomplishments that we're proud of

We are proud of the product we have completed, and although there are some improvements we have not had the time for, we are satisfied with having created a viable product. We recognize that the scope we had for this project, especially as first-time hackathon-ers was very high, and although not all of the goals were achieved we are happy with the progress we made.

What we learned

Throughout this project, we learned a great deal about data scraping and mostly time-series-forecasting, including using recurrent neural networks to solve problems about time-based datasets

What's next for Future Stock Price Prediction

By far the most important thing about this project to me is to improve the prediction algorithm Stock prices have long been studied, with very little success, and any algorithm created to help predict their prices can always be improved for better accuracy. I hope to someday be able to trust this project enough to make real profitable investments based on the results of this program. Notably, in the future I plan to upgrade the recurrent neural network, to one of the higher complexity versions such as GRU or LSTM.

Built With

• java
• nasdaq-api