A Machine Learning Approach that meets Axiomatic Properties in Probabilistic Analysis of LTE Spectral Efficiency

Recent maturity of Long Term Evolution (LTE) has raised interest in assessing how efficiently the cells are deployed. Since Cell Spectral Efficiency (CSE) is a principal indicator of such assessment, we calculate the Cumulative Distribution Function (CDF) of CSE using the measurement data obtained from Rohde & Schwarz TSME which can decode the control channel of the downlink signal. We adopt data processing methods from our previous work for deriving the CDF of CSE. In this paper, we propose a deep neural network model which is modified from the previous one to guarantee that the CDF generated from the model satisfies the probability axioms for any cell configurations. Predicted results from the measured data validate our analysis.