How to get prediction value for instance in weka?

Question

I am working on Weka and should output predicate values ​​(probabilities) for each label for each test instance.

There is an option in the tab as classification in the graphical interface (classify → options → Output, the predicted value) that does this, displaying the prediction probabilities for each label, but how to do it in Java code. I want to get probability estimates for each label after its classification?

Answer 1

The following code takes a set of training instances and displays the predicted probability for a particular instance.

import weka.classifiers.trees.J48; import weka.core.Instances; public class Main { public static void main(String[] args) throws Exception { //load training instances Instances test=... //build a J48 decision tree J48 model=new J48(); model.buildClassifier(test); //decide which instance you want to predict int s1=2; //get the predicted probabilities double[] prediction=model.distributionForInstance(test.get(s1)); //output predictions for(int i=0; i<prediction.length; i=i+1) { System.out.println("Probability of class "+ test.classAttribute().value(i)+ " : "+Double.toString(prediction[i])); } } } 

The "distributionForInstance" method works only for classifiers that can output distribution predictions. You can read it here .

Answer 2

I think I found a solution.

The training set and the test set must be equal: the same title, the same attribute name, the same order. Changes only numbers. And the question arises: why do I need to put the class in a test suite if I do not know it, and that is exactly what I want to get? This method seems to need it, but when you look at classModel.distributionForInstance(dataModel.instance(0)) , it gives you a prediction on your classes with a double array. Thus, you need to put some class values ​​in a test case, and then 'distributionForInstance' gives you the real result for your classes.

Answer 3

From the WEKA GUI, the "Classification" panel → click "More options ..." → Output forecasts → Select "PlainText". Now left-click on “PlainText” and turn “outputDistribution” to “True”.

Please note that this process can be performed in the latest versions of WEKA, for example WEKA 3.8.0.

Yours faithfully,
Martin