Abstract
Blockchain technology is a decentralized ledger that allows the development of applications without the need for a trusted third party. As service-oriented computing continues to evolve, the concept of Blockchain as a Service (BaaS) has emerged, providing a simplified approach to building blockchain-based applications. The growing demand for blockchain services has resulted in numerous options with overlapping functionalities, making it difficult to select the most reliable ones for users. Choosing the best-trusted blockchain peers is a challenging task due to the sparsity of data caused by the multitude of available options. To address the aforementioned issues, we propose a novel collaborative filtering-based matrix completion model called Graph Attention Collaborative Filtering (GATCF), which leverages both graph attention and collaborative filtering techniques to recover the missing values in the data matrix effectively. By incorporating graph attention into the matrix completion process, GATCF can effectively capture the underlying dependencies and interactions between users or peers, and thus mitigate the data sparsity scenarios. We conduct extensive experiments on a large-scale dataset to assess our performance. Results show that our proposed method achieves higher recovery accuracy.