In: Statistics and Probability
Construct a Bayesian Network that functions as a Naïve Bayes Classifier. The naïve assumption lies on the fact that values of attributes are independent conditional on the decision variable. Keep this in mind while creating the network. The construction should include a graph (diagram of the network) and then the conditional probability distribution for the variables.
Naive Bayes is a non linear model. You will see that very clearly in Python or R when plotting the prediction
boundary which will be a very nice curve well separating the non linearly distributed observations.
Suppose, we have data that based on age and salary, how many like to go to office by driving and how many go through a walk to office.
And, simply this data on a graph looks like:-
Now, here comes a new data poiny say X and we have to classify that in which category this new data point lies.
So, we will calculate the probabilities of walks and drives given X and whichever is larger, X will fall into that category.
So, X will be choosen as a point of walks.