A Proposed Class Labeling Approach: From Unsupervised to Supervised Learning
Supervised learning algorithms represent a crucial goal in data analysis fields where the data are groups according to some predefined class labels.
However, many applications in real world come with no class labels in the used datasets, which in turn reduce the possibility of converting such data into knowledge.
This paper presents an approach for automatic class labeling of the objects using clustering-classification approach.
The approach consists of two main phases.
The first one is to group the objects of the dataset into k clusters (unsupervised learning) using k mode clustering algorithm since the available data are categorical type.
The second phase is to convert the clusters received in phase one into class labels with their margins (supervised learning).
An accuracy of more than 85% was received using logistic regression classifier.
The obtained results have been shown high range of accuracy, precision, recall, and f measure for clustering technique and high accuracy for classification.