LTE Network Coverage Hole Detection Technique using Random Forest Classifier Machine Learning Algorithm
| dc.contributor.advisor | Yihenew Wondie (PhD) | |
| dc.contributor.author | Serkadis Bahiru | |
| dc.date.accessioned | 2023-12-14T14:48:27Z | |
| dc.date.available | 2023-12-14T14:48:27Z | |
| dc.date.issued | 2023-10 | |
| dc.description.abstract | The most significant phases for the mobile operator are the performance of cellular networks and the evaluation of quality of service (QoS). Hence, Long-Term Evolution (LTE) network monitoring and measuring are necessary in determining mobile network statistics to optimize the performance of the network in a particular area. Measuring networks is an effective way to qualify and quantify how networks are being used and how they are behaving, and it helps to find voids or uncovered areas that are coverage holes in the LTE network. A coverage hole is defined as a client's being unable to receive a wireless network signal. Traditionally, it is detected through drive tests, but it costs resources and time. In order to identify coverage holes in the LTE network, this study proposed a random forest classifier Machine Learning (ML) approach. The data was gathered from the User Equipment (UE) via minimization drive test (MDT) functionality with low cost, which measures Reference Signal Receive Power (RSRP), Reference Signal Received Quality (RSRQ), and Signal to Interference and Noise Ratio (SINR). The RSRP and RSRQ measure the strength and quality of the reference signal received at the device's antenna, whereas SINR is a metric that assesses the relationship between the targeted signal strength and the total power of all interfering signals and noise. The results obtained show that the employed model's accuracy was 96.3%. However, we could use the hyperparameter optimization (HPO) technique to enhance the model's performance, namely random and grid search cross validation (CV), which increased the model's accuracy to 97.7% and 99.2%, respectively. | |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/953 | |
| dc.language.iso | en_US | |
| dc.publisher | Addis Ababa University | |
| dc.subject | Coverage hole, Random forest classifier, Hyper parameter optimization, minimization drive test, RSRP, RSRQ, SINR | |
| dc.title | LTE Network Coverage Hole Detection Technique using Random Forest Classifier Machine Learning Algorithm | |
| dc.type | Thesis |