One of the goals of an operator is to improve the Quality of Experience (QoE) of a client in networks where Over-the-top (OTT) content is being delivered, as the Internet has now become the `de facto` path for video transmission, with video data representing over 75% of the entire network traffic, and the users' desires to consume more content continuing to show exponential growth. This paper proposes a new approach to improve the clients' QoE by using network awareness models as a way to forecast the number of future video content chunks to be cached. Once cached, these chunks become closer to the client, which mitigates the consequences of suboptimal network performance on bitrate-adaptive HTTP video streaming. Two prefetching approaches are proposed: an Analytic Hierarchy Process (AHP) through a weighted function on the measured quality of service metrics, and a Neural Network to estimate and maximize the user experience without the normal service delivery degradation. Our evaluation shows that AHP provides a suboptimal behavior under constrained bandwidth, while a neural network model is able to infer network behavior and predict an optimized MOS, allowing it to better optimize the caching system and the required prefetching to adequately improve QoE.