Re: Compiling column-wise time series data with DS for data-mining

I would rather go for a master-detail table relationship then. Your master table would contain columns H01 - H25. Your detail table would be pivoted and contain columns:

• H01 (FK to the master)
• Time Series Name ('MN012013', 'MN022013'...)
• Time Series values (most probably 2 columns: timestamp and value?)

You can easily have DS generate and fill both table structures from the unnested data stream in your dataflow.