Traffic Prediction in Telecom Systems Using Deep Learning

Abstract

The deep neural network implementation in this work analyses, evaluates and generates predictions based on the open source big data of telecommunications activity released by Telecom Italia. The deep learning library used for the neural network implementation is Tensorflow which contains many high and mid-level APIs to achieve the functionality. The model uses random data from the test dataset for generating predictions and Estimator API of Tensorflow for building the neural network. Also Adam optimizer is used for optimizing the loss function with the model’s resulting efficiency to be around 98.6–99.8%.