Inspiration
There has been new developments with models like Pragma, this in conjunction with the complexity of FX markets means that traders should be provided with a much information as possible to best inform their decision making.
What it does
This application is trained on the provided dataset using a combination of models from XGBoost and Pragma. Once done there is a web application which takes in live trading data and predicts in real time whether a FX pair is risky with high volatility and whether the pair is a buying or selling opportunity.
How we built it
We built it buy having each team member focus on a specific aspect of the project. So we had one team member focus on the UI; another focus on the XGBoost side; another for Prgama and the final one filled for what was needed.
Challenges we ran into
One problem was the training time of our model, this caused massive bottlenecks in testing and feature development.
Accomplishments that we're proud of
The fact we have a finished MVP that includes all the features we set out to implement.
What we learned
To better plan for the training times of models; more about models like XGBoost and Pragma and how the FX markets work.
What's next for FX Forcast
There's a lot more avenues we can explore to further refine our models but only if we had more time and compute (we might need a H100).
