Web browsing is one of the main Internet services, which has largely evolved over the last years. It now includes many different components (images in various formats, css, Javascript, Progressive Web Apps, etc.) and a web page is composed of many different resources provided by many servers located in various places. Knowing how web browsing content is delivered is crucial for better understanding the Quality of Experience (QoE) offered to end-users. Detecting the factors impacting this QoE is important to improve them but also to be able to predict the quality in order to anticipate degradations to it and propose remediations, which is the main goal of our study. For this, we present an analysis of a measurement campaign we have performed during 6 months (July 2018-January 2019) on the Top 10,000 Alexa websites, which helps to identify the main factors which can impact the quality perceived by end-users. This set of factors has been confirmed by a machine learning process (decision tree), which gives as outputs the set of rules to predict the QoE. The evaluation of our decision tree-based model, on the Top 10,000-15,000 Alexa websites, never accessed before, shows that we can correctly predict the web browsing quality an end-user can get.