algothon_submission

Inspiration

We realized the potentials value of buyer/supplier relationships and corporate governance information in active investments.

What it does

In short, it goes long/ short of US stocks with respective features developed from the above dataset while constraint by limiting factor exposures. It archives a total annual return of 3.55% over 2019.

How we built it

It ingests data from Global Supply Chain Relationships, Supplier Implied Risk Attributes and market data, maps customer-supplier revenue dependencies to the supplier public listed tickers and corporate's governance data to the respective ticker. Then we predict transformed 1-month forward-looking return using an XGBoost model. The model is tested and backtested from the beginning of 2019 until now.

Challenges we ran into

No gpu instance on aws available for free accounts,

Accomplishments that we're proud of

using pandas, lol quantopian and ec2 instances spinning up

What we learned

we are now pandas gurus

What's next for algothon_submission

potentials in enhancing machine learning model (hyperparameter tuning, neural network), control low-volatility exposures in the portfolio