The focus of previous generations of mobile communications, such as First Generation (1G) and Second Generation (2G) networks, was mainly on voice services. Moreover, quality control in those services was less complicated than that of Third Generation (3G) and LongTerm Evolution (LTE) network services. The aim of LTE is to support different services with high data rates and strict Quality of Experience (QoE) requirements for users. Successful mobile communication services can be achieved by managing QoE, improving user satisfaction, etc. The main challenge network operators face is how to monitor and improve Quality of Services (QoS) in real-time to offer differentiated services. Currently, most mobile phone applications, such as multimedia processing, text processing, vision recognition applications, etc., have different traffic characteristics which require several levels of QoS. The streaming of video content through wireless communication networks (in particular, 4G-LTE) is increasing dramatically and gaining popularity in terms of application usage [1]. However, the success of video applications over LTE mainly depends on meeting the user’s QoE needs. Therefore, it is extremely desirable that the network's systems are able to predict and, optimise (if appropriate), video quality. The main goal of this research is to develop new methods to predict the perceived quality of video over LTE accurately and, more importantly, to be able to perform that prediction in real-time and to overcome the drawbacks of the existing methods. Machine learning (ML) and Artificial Intelligence (AI) based methods are utilised in the form of non-intrusive methods with results close to those which can be obtained from intrusive methods. In this research, the QoE prediction is performed by investigating the relationship between the critical video quality and the various QoS parameters, and the selection of the most important QoS parameters is performed to design novel learning models based on Random Neural Networks (RNN). In addition, a new classification of video content types is proposed to develop new learning models for video quality evaluation and optimising video downlink scheduling using Genetic Algorithms (GA) with the aim of maximising the QoE. The proposed QoE models are evaluated and tested using an open source simulator (LTESim) utilising essential video quality characteristics. The outcomes of this work are compared to the relevant learning and optimisation schemes for state-of-the-art analytical approaches such as Adaptive Neuro-Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANN). The results show that our video quality prediction models have higher accuracy, and that our optimisation downlink scheduling can improve performance in terms of QoE.