Service Recommendation Based on Attentional Factorization Machine

With the increasing popularity of SOA (Service Oriented Architecture), a large body of innovative applications emerge on the Internet with mashup (e.g., composition of multiple Web APIs is a representative). Recommending suitable Web APIs to develop Mashup applications has received much attention from both research and industry communities. Prior efforts have shown the importance of incorporating multi-dimensional features extracted from a service repository into their recommendation models. Despite their effectiveness, they are insufficient by simply modelling all these features with the same importance degree, neglecting the fact that not all features are equally useful and predictive. Some useless features may even introduce noises and adversely degrade the performance. In this paper, we propose a novel service recommendation method, which tackles this challenge by discriminating the importance of each feature from data via Attentional Factorization Machine. It endows our model with better performance and a certain level of explainability. In this model, we first extract the valuable features implied in the raw dataset and subsequently transform them to the input format of Attentional Factorization Machine. Then, multi-dimensional information, such as functional similarity, tags, popularity of Web APIs, are modeled by Attentional Factorization Machine to predict the ratings between mashups and services. Comprehensive experiments on a real-world dataset indicate that the proposed approach significantly improves the quality of the recommendation results while compared with up-to-date ones.