system to automatically determine whether a product, such as an
engine, coming off an assembly line is “bad” or “good”. In order to
make the decision, the company observed that three features are
mostly relevant to the quality of the product:
• Wobbly: the engine may be wobbly (which can be detected from a
motion sensor);
• Rumbly: the engine may be rumbly (which can be detected from a
sound sensor), and
• Hot: the engine may be overheated (which can be detected from
a heat sensor).
Each of the above sensors gives a Boolean observation: “true” or
“false”, and an engineer formulates the problem using the following
variables and domains: Decision: B {bad, good} Evidence: W, R, H
{true, false} The engineer wants to use a Naive Bayes (NB)
classifier for making the automatic decision, and builds an NB
classifier which predicts whether a product is “bad” or “good”
(1) Using Naïve Bayes assumption, draw Bayesian network of the
classification model (including all evidence and decision in the
network) [0.5 pt].
(2) Using Naïve Bayes assumption, write formula to calculate
joint probability of the Bayesian network, i.e., P(B, W, R, H),
using symbols B, W, R, H. [0.5 pt]
(3) Using naive Bayes assumption, please write an expression of
posterior probability of P(B|W, R, H) using probabilities of the
forms P(B), P(W|B), P(R|B), and P(H|B) [0.5 pt].
(4) Explain how to use P(B|W, R, H) to determine whether a
product is “bad” or “good” [0.5 pt]
(5) The engineer has observed several engines which recently
came off the assembly line. Based on these observations, suppose a
new engine is observed to have the following feature values:
W=false, R=true, H=false. What prediction will the naive Bayes
classifier make: bad or good? Please justify the answer [1 pt]
- Question 4 3 Pts A Company Intends To Build A Quality Control System To Automatically Determine Whether A Product Su 1 (5.62 KiB) Viewed 20 times