Prediction of Load Time in Web Services for Different Agent Locations and Browsers for Mobile Devices

The paper concerns on the prediction of the load time of Internet resources, depending on the location of the client and the browser used for mobile devices (Android and IOS operating system). An experiment was performed using API of WebPageTest to get and store real data from different agents located in four locations around the world. They monitor and store information about transferring data from 10 sites. The collected measurements were used for making prediction of load time using three different prediction algorithms: the Random Forest, SVM and Linear Regression. The quality assessment of prediction was tested and compared using metrics RMSE, MSE and MAE. Investigations can be applied and implemented in community networks, where time of network selection is important. Results can be also used for performing intelligent internet measurement (IoT), where more and more communication (transmission of data) takes places over mobile devices.