Service Recommendation of Industrial Software Components for Diversified Applications in Industrial Internet
JunJian Liu,Lianglun Cheng,Mingzhe Ni,Tao Wang
DOI: https://doi.org/10.1109/icac61394.2024.10718847
2024-01-01
Abstract:With the development of Industrial Internet, based on service-oriented architecture, it is an important trend to use services to encapsulate IT and OT heterogeneous entities and become reusable, loosely-coupled software components, thus supporting service-oriented componentized software development. To address the different requirements of industrial tasks, it has become a critical issue to improve the accuracy of service recommendation for industrial software components. Existing recommendation models do not give enough importance to semantic information and do not effectively integrate low-order and high-order interactions, ignoring feature domain interactions. Therefore, in this paper, we propose a SAFFM model, which includes Semantic Component and Feature Domain Interaction Component, utilizes Neural Network to encapsulate the high-order interactions, and introduces Multi-Head Attention to evaluate the significance of various feature interactions. For the scenario of us, we designed the service recommendation architecture for industrial software components for diversified applications and performed experiments on genuine datasets. The outcomes distinctly demonstrate the superior performance of SAFFM over state-of-the-art recommendation algorithms, with evaluation metrics such as AUC and log-loss achieving scores of 0.9143 and 0.3734, respectively.