Prediction of optimum internet data transfer speed using artificial neural network

In parallel with the developments in technology, the daily needs of people in their business and private lives are changing. With the spread of the Internet, the need for data sharing between people increases with each passing day. This makes the need for continuous access to the Internet an indispensable part of human life. One of the most common methods for Internet access is Digital Subscriber Line (DSL) technology. DSL is a communication technology that enables simultaneous high-speed data transmission over copper wires used for existing telephones. This internet access is provided through copper transmission lines from time to time due to different reasons or failure to provide the desired internet access speed. While some of these disadvantages require physical intervention, some can be remedied by the management of changing the speed profiles assigned to the relevant user's line due to parameters such as signal to noise ratio and attenuation of the transmission line related to remote access. This thesis is about improving the method of improvement of the problems that can be solved through the system without requiring physical intervention. In the present case, the operators determine which speed profile is suitable for which DSL line. In this study, it is aimed to estimate which speed profile is suitable for the DSL line being studied with different artificial neural networks (ANN) models designed on MATLAB. Proper speed profile estimation is carried out with ANN models designed using parameters such as line attenuation, maximum download speed, signal to noise ratio and output power. With this model, suitable speed profile estimation has been realized with 98.9988% accuracy. More than one ANN model was designed in order to realize the estimation of the speed profile with higher accuracy and in a shorter time and the advantages of the obtained results in terms of speed and performance were discussed.