I have over 500 source files, each of which contains a multivariate time series sequence (same variables though). My objective is to develop a model that can find anomaly by checking the reconstruction error from LSTM (Long Short Term Memory model). Each file, has about 4000 - 5000 records of data. Since the no. of records in each file are not sufficient for LSTM to learn enough features, how do I combine different time series together and run LSTM ? Also, what would be other approaches in this case ?

More Ashutosh Karna's questions See All
Similar questions and discussions