Evaluating Generative Adversarial Networks for Clustering Web Services: A Comparison with Classical ML Models

Web service clustering plays a crucial role in improving service discovery and recommendation systems. This paper proposes a novel approach that leverages Generative Adversarial Networks (GANs) for synthetic text generation, followed by AutoEncoder-based dimensionality reduction for effective clustering of web services. The methodology consists of multiple stages. First, a Seed-to-Text GAN generates synthetic web service descriptions that mimic real-world text. A machine discriminator is then trained to distinguish between real and synthetic descriptions, with periodic human evaluation to prevent model collapse. Next, both real and synthetic descriptions are transformed into embeddings using a pre-trained BERT model, capturing their semantic properties. To enhance clustering efficiency, these high-dimensional embeddings undergo dimensionality reduction via an autoencoder, which learns compact representations while preserving essential structures. Finally, a K-Means clustering algorithm is applied to group web services based on their text descriptions. The proposed approach enhances web service categorization, improving retrieval and recommendation accuracy in dynamic service ecosystems. Experimental results demonstrate effectiveness in producing coherent and well-separated clusters, outperforming traditional clustering techniques.