An Augmentation Experiment in Time Series Modeling

This research is a joint collaboration between machine learning researcher Sangwon Hwang and System1’s Nathan Janos.  It was inspired by the FilterNet time series model developed at System1.  The research in this paper was led by Sangwon Hwang with guidance from Nathan Janos.  In it we test if a CNN can fit and predict time series data with augmentation sub-policies using extremely high variance data from Criteo Labs.

View/Download