An Ensemble Latent Factor Model for Highly Accurate Web Service QoS Prediction

How to accurately predict quality of service (QoS) data is a great challenge in Web service selection or recommen-dation. To date, a latent factor (LF)-based QoS predictor is one of the most successful and popular approaches to address this chal-lenge as its high efficiency and scalability. However, current LF -based QoS predictors are mostly developed on inner product space with an L2 norm-oriented loss function only, thereby they cannot comprehensively represent target QoS data's characteris-tics to make accurate prediction as inner product space and L2 norm have their respective limitations. To address this issue, this study proposes an ensemble LF (ELF) model. It has three-fold ideas: 1) two kinds of LF models are developed as QoS predictors on inner product space and distance space, respectively, 2) both of these two QoS predictors adopt an Ll-and-L2-norm-oriented loss function, and 3) building an ensemble of these two QoS predictors by a weighting strategy. By doing so, ELF integrates multi-merits originating from inner product space, distance space, L1 norm, and L2 norm, making it achieve highly accurate and robust QoS prediction. Experiments on a real-world QoS dataset demonstrate that the proposed ELF model outperforms state-of-the-art QoS predictors in predicting the missing QoS data.