PCA modeling of extraction and selection of variables influencing LTE network delay in urban mobility conditions

Abstract

In addition to network throughput, which defines the number of bits per second that can be transmitted, another primary measure of network performance is latency, or delay, which defines the time spent sending one bit from one computer to another in the network. In this paper, the main goal of the research is to model the efficient management of delay performance in a 4G Long Term Evolution (LTE) network on a defined section of the trunk road with the assumption that key performance indicators (KPI) can be effectively analyzed with the created predictive classification models. These models also enable the assessment and prediction of the fulfillment of sustainable urban mobility plans. Using Principal Component Analysis (PCA), the space of 17 input variables is reduced to four extracted components. Several models based on different machine learning techniques are created using the automatic modeling method, and the final solution is selected according to the criterion of maximum classification accuracy, interpretability, and complexity. A classification model based on Logistic Regression was chosen as the final solution.