I participated in my second data science competition last month and it was again a worthy experience.
My team, composed of myself and two friends, competed with 131 teams from all over the world at Capital Fund Management’s 2019 supervised machine learning challenge, hosted at ENS’s Challenge Data platform.
The goal was to predict sign accuracy using anonymized US stock returns data, a highly stochastic dataset with 1 million lines and 74 columns (training + test).
The group split between different tasks such as feature engineering, essaying neural network architectures (my part) and AutoML frameworks, and applying traditional ML methods like GBMs.
Our best submission was a CNN-based one that unfortunately overfit the training data a little, which gave us the final ranking of 15.
We did this in less than a month. I believe that we will participate again this year and we're going to go even further since we have more time.
The first place wins an iPad. I never had one and, oh boy, I want it.
After winning in 2020, all three of us then can become quant traders and be filthy rich for the rest of our lives. Finance is awesome!
Stay tuned next year to find out if the plan goes accordingly.