DMX 102 - SSAS Time Series with Multiple Attributes

This post looks at extending the time series prediction so that multiple values can be predicted. In the last post (DMX 101) only a single value was predicted and while this can be useful, the approach has obvious draw backs because there is often an expectation that different classes have different patterns. Consider seasonal fashion. We expect that seasonal products will have a distinct life-cycle and to group products (from multiple seasons) would distort a specific seasons sales.