Personalized QoS Prediction for Web Services Using Latent Factor Models

Recommending the suitable Web service is an important topic in today´s society. The critical step is to accurately predict QoS of Web services. However, the highly sparse QoS data complicate the challenges. In the real world, since QoS delivery can be significantly affected by some dominant factors in the service environment (e.g., network delay and the location of user or service), Web services which are published by the same provider usually have the similar fundamental network environment. These factors can be leveraged for accurate QoS predictions, leading to high-quality service recommendations. In this paper, we expound how Latent Factor Models (LFM) can be utilized to predict the unknown QoS values. Meanwhile, we take the factors of provider and its country into consideration, which imply the latent physical location and network status information, as the latent neighbor for the set of Web services. Hence, the novel neighbor factor model is built to evaluate the personalized connection quality of latent neighbors for each service user. Then, we propose an integrated model based on LFM. Finally, we conduct a group of experiments on a large-scale real-world QoS dataset and the results demonstrate that our approach is effective, especially in the situation of data sparsity.