Hi Yong June
ThingWatcher works a little bit as follow:
- - during calibrating mode, a training model is created that represent the correct state of work
- - during monitoring, ThingWatcher looks at the last few data to predict the value of the new incoming data.
- - it then compare the predicted values with the actual value received and flag it as anomalous or not depending if it fits the prediction or not.
The lookback window represent the amount of data ThingWatcher will use to make the prediction in step 2 above.
A usual range is max lookback 16 , min lookback 8, but this may need tuning.
Note that in the latest release of ThingWatcher it is NOT recommended to set the lookback window, instead ThingWatcher will automatically test a few and keep the optimal one. See the Help Center.
Since you are talking about the Excel tool, I would assume you are using an old beta release of ThingWatcher (The Excel tool no longer exist in latest release), I would recommend that you use the latest release to get the best results.
Hope this helps