Saturday, October 01, 2011 9:13 AMI'm building a Time Series model with data stored at a high level of time granularity. The time dimension uses DateTime as key with milliseconds precision. There're about 500000 records in total. What is the best approach to build a Time Series prediction with minute granularity, calculating minute averages? Should I create MDX members for averages for inputs, or is it more efficient to modify the underlying relational data?
Wednesday, January 11, 2012 4:25 PM
Please go thorough the below links, It will be useful for your requirement
Wednesday, January 11, 2012 5:03 PMAnswererYou can create a cube that stores aggregates for each minute and build time series model on the cube.
Tatyana Yakushev [PredixionSoftware.com]