Component page in Cerebro
Check a component's risk, its 7-day prediction versus what happened, the most influential variables and its alerts.
On this page
A component's page in Cerebro brings together everything Cerebro knows about it. You open it from Cerebro → Devices, from the Highest risk list on the Summary page or with View details in Cerebro on the component's page.
At the top
The component's type and name, and the state of the model for its type (for example, Training or Anomaly mode) with a short explanation.
View page: opens the component's regular page.
Disable Cerebro: only if you can operate.
Indicators
Current Level.
Health (1 − 24 h risk) and the risk at each horizon, when there is a percentage.
Unusualness vs. its history, in anomaly mode.
The model's mode and the time of the last prediction.
Prediction versus what happened
A 7-day chart with the predicted risk and, marked on top, the failures and alerts. At the top you'll see the summary:
How many failures there were and how many were anticipated.
How many alerts there were, how many were correct and how many were false alarms.
The High-risk threshold used for this component, and whether it is learned for this component or shared by the type.
If the model already calculates percentages, choose the Horizon on the right.
Most influential variables and signals
Most influential variables: what influences the model for that component type the most, with its percentage of the total weight.
Signals: what is driving this component's risk today (▲ raises, ▼ lowers) and the disk forecast if applicable.
Model
It explains in simple words how the model did in its test, for example “It anticipated 70 out of every 100 failures in the test period, about 6 h in advance”, how many false alarms it gives per week compared with the current rules, and when the next training is.
Predictive alert history
All of the component's predictive alerts, with their answer. If an alert is open and you can operate, you can answer it from here. See Predictive alerts.