I have multivariate information, about 9 dimensions of time series collected from different sensors. I want to combine or consolidate these signals into ONE SIGNAL time series to represent that event. I will then use that single time series to do anomaly detection. What are some recommended methods (mathematical or machine learning), besides PCA, to reduce my 9-D time series to 1-D? The 1-D single time series should effectively capture the nuances of that time period from all the sensors. And then some classification or regression methods can be used afterwards. I want to avoid using PCA and explore other available methods. Are there software packages that can create single-unified time-series models from multivariate time series signals? Continue reading...
