A Novel Hybrid Framework for Unsupervised Clustering Integrating Artificial Neural Network Architectures

Abstract

Unsupervised clustering remains a fundamental challenge in machine learning due to the absence of labeled data and the complexity of discovering intrinsic data structures. Traditional clustering techniques such as K-means and hierarchical clustering often suffer from sensitivity to initialization, scalability limitations, and inability to capture nonlinear relationships. In contrast, artificial neural networks (ANNs) have demonstrated strong capabilities in modeling complex patterns but require structured integration to function effectively in unsupervised environments. This study proposes a novel hybrid framework that integrates artificial neural network architectures with classical clustering algorithms to enhance clustering accuracy, robustness, and adaptability. The framework leverages neural feature transformation, adaptive similarity learning, and iterative refinement mechanisms to improve cluster formation. Experimental evaluation is conducted using benchmark datasets, including Iris and Wine datasets, demonstrating superior performance compared to conventional methods. The results indicate significant improvements in cluster cohesion and separation. The proposed approach contributes to the advancement of unsupervised learning by bridging the gap between neural computation and clustering methodologies, offering a scalable and interpretable solution for complex data analysis.