Timeseries prediction with label being next timeseries element instead of own column?


I’d like to predict values of a timeseries. For that I want to load a csv file containing sampled data (7 features per message).

I’m using CSVVariableSlidingWindowRecordReader with a maxLinesPerSequence of N. I’d now like to get datavec / dl4j to automatically use the message at X + N + 1 as the label, where X is the current start of the sliding window. Is there a way to achieve this or do I manually have to produce create a second csv containing the data that is to predict?