Automatic Group Labeling Using Attribute Information Gain Filters and Unsupervised Learning

Automated group labeling is a challenging task in machine learning, as it requires interpreting groups through labels that accurately represent their defining attributes. This paper proposes a method for automatic group labeling using attribute information gain filters. Our approach involves applying unsupervised learning algorithms to form groups and selecting relevant attributes using filters to generate labels. Additionally, we employ a discretization method to aid in calculating the variation of data values. We tested our method on four datasets available in the UCI repository (Iris and Wine) and achieved an average accuracy rate of 83.66%, with an average standard deviation of 4.98%. Our results demonstrate the effectiveness of our approach in automatically generating accurate and meaningful labels for groups.