I have a M multivariate time series data, by multivariate I mean that a time series is represented by more than one variable that varies in time (see example image for simulated data). All have the same size. I want to build a classifier trained on K class (eg. all time series data belongs to A, B or C class).
Is there a straightforward implementation of this in R, specifically, as the regular classification approaches (e.g random forest, SVM) will ignore the dependent data and give different predictions within the same time series. I have an intuition how this could be solved, e.g. using some ensemble classification, or concatenating time series into a univariate vector, but I have a feeling there is a better approach for this that doesn't require me to reinvent the wheel. I also know that KNN and DTW approach could in theory work, but not sure how they get around these issues above (e.g. the multivariate problem)
Appreciate any pointers and references
