Market Vision

Inspiration

We didn't just want to build a standard sentiment analyzer - instead, we wanted our bot to properly understand the statements!

What it does

Our approach for the sentiment analyzer involves a Recursive Neural Network (RNN) that takes the message or post as an input and outputs a vector containing numeric values that correspond to the available stocks. If the value at the index of a certain stock is negative or positive, we either sell or buy the stock respectively according to the breaking news. Basically, instead of just evaluating the sentiment of a news headline, we calculate its relevance to the individual stocks.

How we built it

We used the platform Optibook to visually check market movements and its API to gain access to the exchange, both provided by Optiver. Our neural network uses the tensorflow library.

Challenges we ran into

Our AWS instance had some errors we couldn't solve, so we had to get new credentials and lost some time. Furthermore, it's quite tough to correctly train a neural network with so little training data.

Accomplishments that we're proud of

Not only making losses, and having quite a lot of success in comparatively little time.

What we learned

How to get stonks which specifically for us is learning more about trading at a stock exchange and getting a better understanding of NNs.

What's next for this project

Getting a bigger training data set and refining the architecture of our NN.

Team Number

We participated on optibook as team-061

Running the Project

All of our changes build directly on the Basic Quoter.ipynb file. Running it will activate all of our additional programs.

Built With

• amazon-web-services
• cloud9
• optibook
• python
• tensorflow