Unsupervised Learning Aided by Hierarchical Analysis in Knowledge Exploration

Unsupervised learning plays an important role in knowledge exploration and discovery. Two basic examples of unsupervised learning are clustering and dimensionality reduction. In this paper, we introduce an improved model for clustering based on a hierarchical analysis method. In our model, there are three main steps. In the first step, we use a structural clustering model to find qualitative patterns from a given dataset. Then, the second step applies a quantitative-based clustering algorithm to find quantitative patterns from the dataset. The third and the last step generates hybrid patterns by combining the patterns obtained from the first two steps based on a certain criterion so that deeply hidden relationships can be extracted from the dataset. In this paper, we also discuss the results of our experiments with the proposed model and algorithms on longitudinal medical records