In an HTTP streaming framework, continuous time quality evaluation is necessary to monitor the time-varying subjective quality (TVSQ) of the videos resulting from rate adaptation. In this paper, we present a novel learning framework for TVSQ assessment using linear regression under the Reduced-Reference (RR) and the No-Reference (NR) settings. The proposed framework relies on objective short time quality estimates and past TVSQs for predicting the present TVSQ. Specifically, we rely on spatio-temporal reduced reference entropic differencing for RR and on a 3D convolutional neural network for NR quality estimations. While the proposed RR-TVSQ model delivers competitive performance with state-of-the-art methods, the proposed NR-TVSQ model outperforms state-of-the-art algorithms over the LIVE QoE database.