Granger Components Analysis

Discovering the structure in time series datasets is a fundamental endeavor of data science and particularly forecasting applications. I have developed an unsupervised learning technique that employs the notion of Granger causality to blindly learn pairs of signals with strong temporal dependencies.