Adaptive Weighted Learning for Unbalanced Multicategory Classification

Abstract

In multicategory classification, standard techniques typically treat all classes equally. This treatment can be problematic when the dataset is unbalanced in the sense that certain classes have very small class proportions compared to others. The minority classes may be ignored or discounted during the classification process due to their small proportions. This can be a serious problem if those minority classes are important. In this article, we study the problem of unbalanced classification and propose new criteria to measure classification accuracy. Moreover, we propose three different weighted learning procedures, two one-step weighted procedures, as well as one adaptive weighted procedure. We demonstrate the advantages of the new procedures, using multicategory support vector machines, through simulated and real datasets. Our results indicate that the proposed methodology can handle unbalanced classification problems effectively.

Three Categories Customer Churn Prediction Based on the Adjusted Real Adaboost

“…Then, we give the following: W e t j j = 1/ j ∧ e j . A study gave the evidence pointing to the efficiency of this weight form is the results (Qiao and Liu, 2009). They compared the different forms of the weight and showed it is a feasible solution.…”

“…Error rate is the traditional performance criterion, that is, the percentage of incorrectly classified observations in the validation set. However, taking the extremely imbalanced data into account, the traditional error rate measure is inappropriate for such rare event (Qiao and Liu, 2009); because such a rule pays more attention to the majority of the non churners, it may not isolate the potential churners. Another disadvantage is that error rates do not use the numerical values of the scores, whereas these scores may contain relevant information.…”

“…This article borrows the tree-based ensemble idea from the Real Adaboost and the adaptive weighted learning study from Qiao and Liu (2009). Real Adaboost put higher weight on the misclassified training cases.…”

Regularized receiver operating characteristic-based logistic regression for grouped variable selection with composite criterion

“…If we use the overall error rate as the prediction evaluation criterion, we may get a result biased to the majority class. [8,9] The second issue is the cut-point in prediction. Although a probability of .5 is usually adopted as the default cut-point in the logistic model, it may lead to a biased prediction to majority class for the imbalanced data.…”

Utility-based weighted multicategory robust support vector machines

“…In that case, the resulting classifier tends to sacrifice the minority classes and try to classify the training points in majority classes correctly. Sometimes the classifier may misclassify all points of a minority class but still give high overall classification accuracy (Qiao and Liu, 2009). Therefore, unequal cost assignments on different types of misclassification are needed.…”