CODEWITHSUNDEEP
machine-learningclassificationweka
user27379
user27379

Reputation: 47

Combination of classifiers weka

I have built three classifiers based on 107 instances,11 features and two classes each stage . Weka used as machine learning tool.

  1. First classifier predicts class 0 and class 1-2-3. ( all 107 instances are used for training and testing in cross-validation method)

  2. Second classier predicts class 1 and class 2-3.( The instances with class 0 removed for training and testing)

  3. Third classifier predicts Class 2 and Class 3. ( The instances with class 1 removed for training and testing)

Randoforest is applied to each classifier. Does anybody know how can I combine these three classifiers?

Upvotes: -1

Views: 1130

Answers (1)

lejlot
lejlot

Reputation: 66775

It seems a bit werid pack of classifiers, but the most obvious solution would be to build a "meta classifier" on top, which will decide what is the real class. Consider training a classifier on data of form:

Inputs:

  • ourput_of_classifier1(x) (in unary format)
  • ourput_of_classifier2(x) (in unary format)
  • ourput_of_classifier3(x) (in unary format)

Output:

  • label of x

so you transform your original data into representation consisting predicted labels, and train new classifier on such data.

The simplest possibility would be to train a Naive Bayes to build conditional probabilities in form of P(final_class=Y | classifier1(x)=y1, classifier2(x)=y2, classifier3(x)=y3)

Upvotes: 1

Related Questions