Menu
I am taking classes on intro to AI,and the teacher mentioned some point that for the classifier ZeroR,the accuracy under ZeroR is a helpful baseline for interpreting other classifiers. I searched online about this but still couldn't get my head around it,could anyone give some idea on what that means please,thanks in advance.
929
A classifier utilizes some training data to understand how given input variables relate to the class. In this case, known spam and non-spam emails have to be used as the training data. When the classifier is trained accurately, it can be used to detect an unknown email.
A good classifier will reduce the number of errors smoothly when the threshold is applied which will lead to a rising upper curve. In the same way the correct items will be diminished producing the reject set. This is shown in the schematical graph below with the three sets of items, the Errors, Correct and Rejects.
Now, let us take a look at the different types of classifiers: Perceptron. Naive Bayes. Decision Tree.
I think the line of reasoning runs as follows: A ZeroR classifier simply assigns every value to the most common class (as found by examining the training data). This means that if your data is 55% class A, 10% class B, 5% class C etc then ZeroR will get 55% right. If your data is 33% class A, 31% class B, 28% class C etc then ZeroR will get 33% right.
Save from randomly selecting classes, this is pretty much the dumbest classifier you can get and so you can measure other classifiers by how well they do compared to this minimal level of performance. Given a certain data set, you can use ZeroR to find out what the minimum performance is you may expect.
167
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
