The analysis conducted in this repository is currently conducted on $TSLA options.

We have so far attempted Random Forest Regression, Gradient Boosted Regression, Kriging, Nearest Neighbor, directly imported from the scikit learn python package and the pykrige package.

https://github.com/siddhantdubey/volatility/blob/master/Graphics/FitImages/forestregression.png?raw=true)

The following output is fairly bad, this is most likely due to poor implementation. Kriging is a technique taken from geo-statistics.

This uses the Multi Layer Perceptron method, the following was done with a hidden layer size of 400, trained over 450 epochs.