How User Engagement Metrics Ameliorate the Web QoE?
- PDF / 1,399,409 Bytes
- 20 Pages / 439.37 x 666.142 pts Page_size
- 86 Downloads / 189 Views
How User Engagement Metrics Ameliorate the Web QoE? Nawres Abdelwahed1 · Asma Ben Letaifa2 · Sadok El Asmi1 Accepted: 11 November 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract User engagement in general is an interpretation of an individual’s reaction to an offer (a service, a product, an application). We can determine a person’s degree of commitment directly by interacting with him or by observing his behavior. To measure the user engagement some actions can be interpreted such as clicks on links, comments and download of documents. In an other hand, we can measure the user satisfaction towards a web page using web QoE metrics such as Time to First Byte (TTFB), onLoad (Load Time) and Above the Fold (ATF). These metrics are commenly used to predict web QoE using Machine Learning (ML) algorithms. By comparing the two groups of metrics (user engagement and web QoE) we conclude that user engagaement metrics are closer to user desires, that’s why in our work we focus on user engagement to predict user’s satisfaction by predicting the Mean Opinion Score (MOS). Existing works dont use user engagement metrics in datasets to predict MOS, that’s why in this paper, we propose to focus on user engagement to predict our MOS. To do so, we use the help of an existing dataset whose parameters are web QoE metrics. At first, we select metrics that have a direct relation with user enagement. Then, we refine their coefficients to obtain the best combination that gives a MOS very close to the real one (expressed directly by users). After that, we add a new column to the existing dataset that containes the new obtained engagement parameter. Finally, we apply different ML algorithms on our new dataset to predict the MOS and we conclude that decision tree is the best in our case. Keywords User engagement · Web QoE · ML · Deep learning
* Nawres Abdelwahed [email protected] Asma Ben Letaifa [email protected] Sadok El Asmi [email protected] 1
COSIM LAB, SUPCOM Tunisia, University of Carthage, Carthage, Tunisia
2
MEDIATRON LAB, SUPCOM Tunisia, University of Carthage, Carthage, Tunisia
13
Vol.:(0123456789)
N. Abdelwahed et al.
1 Introduction The 90s mark the beginning of the use of the web. And since, the number of websites and web users have increased exponentially. In January 2018, more than half of the world’s population access the Internet, there are 4.02 billion Internet users, a penetration rate of 53% [1]. At first, web pages were static, composed of only static content such as text and images. Later, it evolved to a dynamic and complex content, which is the case of actual web pages. These latter, include so many objects and scripts, distributed on several servers hosted in different domains [2]. The complexity of modern web pages has evolved the web architecture, in order to improve the quality of service (QoS) of the end user and thus the quality of experience (QoE) : introducing the distribution network of content (CDN) and different protocols such as HTTP2 [3]
Data Loading...
