Inspiration
We wanted to make money! In order to be a smart investor, we need to predict the market better than anyone else. As esteemed physicists and computer scientists, we had the thought of how quantum computers could help us do that. If quantum algorithms present an advantage in market prediction, we can gain the upper hand against others!
What it does
Compares the performance of several quantum models over selected features if the market to classical counterparts. These models act as feature augmenters: they take in features and create new ones highlighting dependences between features. This is fed into the prediction module, which takes the augmented features to perform the prediction. To have a controlled experiment, the only thing that changes between tests are the feature augmenters.
How we built it
We first settled on the architecture described above with the feature augmenter and the predictor. We first built the predictor as a simple linear regression model and then built the various different feature augmenters to test. Training pipeline: train the feature augmenter and predictor jointly using a "sliding-window" technique for the S&P500 data.
Challenges we ran into
We were not able to get the S&P500 data to work for the quantum augmenter due to the hardware not working. Since the S&P500 data has many more features (15) than the synthetic data (4), simulating it would be extremely slow and so running it on the actual hardware will be.
Accomplishments that we're proud of
We were able to successfully demonstrate that the quantum augmenter doesn't seem to outperform the classical augmenter for some small synthetic data that simulates the complexity of the stock market. While we were not able to show this for the actual market, we may be able to infer what results we might get because of this.
What's next for QFinance
Getting the hardware to work so that we can simulate the real data!
Log in or sign up for Devpost to join the conversation.