QoE Model for Addis Ababa LTE Web Browsing Service Using Neural Network Approach

In order to address the customer's satisfaction, mobile operators try to find out what the customer needs and what quality makes the customer satisfied. The customer satisfaction can be measured or estimated by Quality of Experience (QoE) measurement. Its estimation and measurement is important to identify the network problems, to understand causes and contributing factors. Web browsing is one of the widely used application on Long Term Evolution (LTE) networks. Therefore, it is essential for service providers to ensure a better QoE on web browsing service. Web QoE can measure the user satisfaction by subjective or objective measurement. Subjective test suffers from some drawbacks, such as it has high cost in terms of time, money, and manual effort and also cannot be used for real-time QoE evaluation. In Ethiopia only subjective measurement is used, to know the level of customer satisfaction. Due to that, the company is exposed for high expenses and also can not perform the real time measurement of QoE. To overcome the problem on subjective test, this thesis developed a web browsing QoE model, using Neural Network algorithm that is implemented in matlab software. The model takes the following QoS metrics as input parameters: page response delay, page content browsing delay and page download throughput. The model map these metrics to QoE interms of Mean Opinion Score (MOS). The model performed an estimation of QoE with a Mean Square Error (MSE) of 0.002 and correlation of 97.2%, relatively to the target QoE. As the result indicates, the estimated and measured QoE values are highly correlated. And the error between them is very low. So, this model can be used for estimating the web browsing QoE for the mobile operators, to get objective measurement advantages. Also, it can be used for operators to identify the network factors that most influence the web browsing QoE.