🧥

Veil is used to predict market data given short-term market events.

Reasoning

Wanting to be quants, we don't really have the opportunity to work with market data, and the Kaggle comp was probably our best chance at doing so. We would like to get hands on practice with data prior to becoming a quant.

Approach

• We pre-processed data, moved towards cleaning and prioritizing all quantitiative data that seemed impactful towards prediction. We also removed all NaN and imputed missing values.
• We trained a baseline XGBoost model to run predictions through backtesting.
• We then refined our model and tested against each other per weight.
• We moved to LSTM to also run predictions and provide insight to our refined model.

Limitations

• Our PCs really sucked (2022MBP; RTX3060)
• I ran out of ram
• I got an overheat warning
• My battery died

Intended contents

We have 10 parquets of data (each 500mb -> 10GiB) and cannot be uploaded to the repository due so.
Notebooks should be enabled for all to use, given that you have parquet data.

Reference

Jane Street