The Catch 22 Feature Set for Time Series Analysis

i spent some years collecting thousands of different time series analysis methods that turn a time series to a single real number. i ended up with over seven thousand of these numbers from different areas of time series analysis. And so we took those seven thousand took those that were the most robust across different types of data, and then applied them to about a hundred different classification problems. We wanted to kind of reduce the redundancy and keep as much information in this feature based representation of a time series as possible.