Semi-Supervised Clustering Using Genetic Algorithms

A semi-supervised clustering algorithm is proposed that combines the benefits of supervised and unsupervised learning methods. Data are segmented/clustered using an unsupervised learning technique that is biased toward producing segments or clusters as pure as possible in terms of class distribution. These clusters can then be used to predict the class of future points. For example in database marketing, the technique can be used to identify and characterize segments of the customer population likely to respond to a promotion. One benefit of the approach is that it allows unlabeled data with no known class to be used to improve classification accuracy. The objective function of an unsupervised technique, e.g. K-means clustering, is modified to minimize both the within cluster variance of the input attributes and a measure of cluster impurity based on the class labels. Minimizing the within cluster variance of the examples is a form of capacity control to prevent overfitting. For the the output labels, impurity measures from decision tree algorithms such as the Gini index can be used. A genetic algorithm optimizes the objective function to produce clusters. Non-empty clusters are labeled with the majority class. Experimental results show that using class information improves the generalization ability compared to unsupervised methods based only on the input attributes. The results also indicate that the method performs very well even when few training examples are available. Training using information from unlabeled data can improve classification accuracy on that data as well. ∗http://www.math.rpi.edu/ bennek