Alpha from Clusters

Traditionally, people have been looking at mean and variance for alpha signals. Oftentimes these are based on just 1 feature.

With the rise of machine learning, more interesting and unconventional signals can be generated from a dataset.

The hypothesis of our project is to use clustering methods to generate state of the art alpha signals described by the clusters’ attributes, adding new and more complex ways of describing a dataset on top of the conventional statistics such as mean and variance.

This project investigates this hypothesis by looking at Social Media Analytics from Quandl to generate alpha signals for NVDA price.

Built With

• python