Forums

> Are the rca_prior files inconsistent?

It seems like the rca prior for the initial sample dataset was representing alarms, but the datasets 1, 2, and 3 are referring to devices in the graph?

1. The description on the website states an example with a simplified screenshot as three alarms (A, B, C) and the root cause being alarm A.
2. In line with 1, the example dataset had a topology of shape 45x45 representing 45 devices. The causal prior (and true graph) have a shape of 59x59 and the root cause for the example (6, 51) is 51. This suggests that the indices are referring to the alarm_ids.
3. For the real datasets, the indices seem to correspond to network ids:
* For the first dataset the topology matrix is 35x35 and the indices range from 0 to 34, whereas the causal prior has shape 39x39
* For the second dataset the topology matrix is 42x42 and the indices range from 0 to 41, whereas the causal prior has shape 49x49
* For the third dataset the topology matrix is 39x39 and the indices range from 0 to 38, whereas the causal prior has shape 31x31 -> This is ruling out that the number refer to the alarm_ids and like represent device_ids.

Could you please verify that this was intended and clarify how this data is intended to be used?

Thanks!