Stand Alone and Clustered Base Stations Approaches for AI Based Congestion Prediction on ORAN RIC Layer

Abstract

5G/6G will rely heavily on the architect concept of Open Radio Access Network (oRAN). In a such open environment, a near real time processing and decision-making capabilities are an essential requirement. Artificial Intelligent, more specifically, Machine learning is the selected technique to be used in performing the needed processes. In this paper, we are presenting the first look at the performance of selected ML techniques in order to practically test their performance using real Data set gathered in cooperation with the Data analytic Dept. at the Saudi Communications Company in Saudi Arabia. The Data set consist of an hourly measurement of certain network’s parameters for the time span of on year and for different seven base stations all are located in Riyadh, Saudi Arabia. Two approaches were studied. in the first one a dataset from one BS only is used for training and testing for different techniques. While, in the second approach, all data from the seven base stations were used instead. Results show that: for a supervised classifier techniques, the Decision tree technique performs the best among the selected techniques with an accuracy of 0.96 and 0.91 for single and all base stations, respectively. For the supervised regression approach the accuracy level was found to be 0.98 and 0.97 for single and all base stations, respectively.

Keywords

• 5g 6g ric ai machine learning congestion prediction