The importance of increasing market shares in today's competitive world makes almost all services convince a majority of people to invest and consume their services. Video services are one of the most competitive services that have a significant impact on users' consumption. Additionally, users' consumption has a strong direct relation with users' satisfaction. Nowadays, user satisfaction, which is also called quality of experience (QoE) in specialized terms, has attracted the attention of large companies and a wide range of research. Improvement of the quality of the experience of video has so many aspects, for instance, category and content of video, visual attractiveness, and video playback condition. Deliberating the aspects mentioned above needs a good cognition level of users' interests and behaviour, and due to the continuous increase in the number of users and the importance of attracting new users with different interests, prediction is one of the main tools for improving the quality of experience. Predicting QoE helps video streaming services and content providers to invest and produce the content that probably achieves the most satisfaction of users and this means quality of experience will increase automatically. In this research, a prediction system which is based on machine learning has been implemented on a popular video service that outperforms similar works that have the same research scopes. In this scope, performance evaluation is based on some numerical metrics that indicate the accuracy and correlation of actual values and predicted values (e.g., root mean square error, linear correlation coefficient). For standardization and making it comparable, a typical target value is used called mean opinion score (MOS). MOS, which is in the subjective scores category, is a value gathered directly from users who watch a video, and it is in the 1 to 5 range (1 indicates the lowest satisfaction, and 5 indicates the highest satisfaction). Additionally, the implemented system finds the best categories that can help content providers improve the quality of experience independent of video playback conditions and any kinds of impairments. Also, a module has been implemented to make any video more attractive based on visual attractiveness. Visual attractiveness has so many kinds of aspects that can impact on human visual system (HVS). The human visual system can be influenced by sharpness, colourfulness, clarity, and brightness. Alternatively, the implemented module has a significant impact on these aspects.
