Automated Stock Software

Inspiration

One of our members had had a similar idea in their head for quite some time, but had never managed to build it. They suggested it to the group, and, being broke college student, all members jumped at the opportunity, especially as it was far superior to our previous idea of an electric shock alarm clock.

What it does

"Automated Stock Software" uses Google's Tensorflow and Google cloud's natural language processing libraries, as well as historical stock data from Alpha Vantage to predict the closing prices of stocks on a day to day basis. It pulls live data from Twitter and uses Google's Cloud Natural Language API to judge the overall confidence in a particular stock, as well as using long-running historical open, close, high, low, and volume data from the stock's trading history.

How we built it

We used a ticketing system to distribute tasks in the most efficient way possible, assigning the machine learning back-end to the group members with the most machine learning experience and allowing the other group members to focus on quickly developing the interface with the outside world in order to provide data to the brain of our project as quickly as possible.

Challenges we ran into

The Twitter API we began the project with limited the historical data we could use too severely to allow its use, so we had to rewrite much of an open source 3rd party API to work with our system.
None of our members had sufficient experience with Tensorflow prior to Brickhack, so a few members taught it to themselves over the course of the weekend. The only free stock data APIs provide data at a limited resolution so we had to work with lower quality data than we had originally hoped for. Google authentication restrictions limited the portability of our application through Docker containers.

Accomplishments that we're proud of

We as a group made much more progress through efficient teamwork than most individual members though possible.

What we learned

Every member became more comfortable with the use of arbitrary APIs, as well as common tools in the industry, such as Docker.

What's next for Automated Stock Software

We plan to continue developing Automated Stock Software to the point we feel comfortable with it deciding on real life trading.

Unfortunately, we cannot provide a github link due to the contamination by private API keys and other sensitive information.

Built With

• alpha-vantage
• google-cloud
• python
• tensorflow