Hi,

I build up a Bayesian network using structure learning algorithm- hill-climbing using R package bnlearn. I am analysing the TEP dataset which contains 55 variables, one being fault_number(containing the fault type) and rest are process variables(containing some sensor data).

My question is after building the network, how can I know the worthiness of this network.

One way I tried is to predict the fault type given process variables. But I was wondering is there any other way to figure out the network I made is good enough(trustworthy).

My ultimate aim is to figure out the probability of any fault if i change one or combination of process variables by some unit.

Thanks