Outdoor Propagation Pathloss Model for UMTS Networks using Feedforward Backpropagation Neural Network: Case of Addis Ababa, Ethiopia
| dc.contributor.advisor | Yalemzewd, Negash (PhD) | |
| dc.contributor.author | Yosef, Mekonen | |
| dc.date.accessioned | 2020-03-11T10:51:16Z | |
| dc.date.accessioned | 2023-11-04T15:13:16Z | |
| dc.date.available | 2020-03-11T10:51:16Z | |
| dc.date.available | 2023-11-04T15:13:16Z | |
| dc.date.issued | 2019-12 | |
| dc.description.abstract | The ubiquity nature of mobile networks, growth in technology, and innovations in mobile services will attract more users with the growing expectation and satisfaction. To accommodate the increasing number of users’, operators are investing more in infrastructure and service delivery. In order to get the right revenue with the appropriate investment, the network planning and optimization work has to be done properly. In this regard, propagation pathloss is one of the main inputs for the planning and optimization and it has to be predicted as accurate as possible. Prediction of the propagation pathloss can be done using different models. These models are deterministic, empirical, and statistical models. From these models, the empirical pathloss models, such as, COST-231, ECC-33, Stanford University Interim (SUI) are more commonly used. Different environments own different model and accuracy. These models have got different accuracy and they are modeled for different environment. When they are applied for the environment other than the area they are modeled for, they lose their accuracy. Hence, searching for a better prediction model is essential. To this end, neural network-based model is one of the better solutions to empirical and deterministic models for predicting the propagation pathloss. The dataset is collected from measurements through a drive-test and from low level design documents. Then, it is preprocessed before used in order to train and evaluate the network. The performance evaluation is done with metrics, such as Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and the coefficient of determination (R 2 ) and Regression coefficient (R). The result of the tthesis shows that the neural network-based model has improved the pathloss by 6.2 dBm, 4.85%, 8.98 dBm, 0.44 and 0.53. The achieved result for the empirical models considering values of MAE, MAPE, RMSE, R and R are 10.57 dBm, 8.34%, 14.36 dBm, 0.38 and 0.14 respectively. On the other hand, the proposed approach achieved a value of 4,37 dBm, 3.49%, 5.38 dBm, 0.82 and 0.67 using the test dataset. To this end, the neural network-based model best fits the pathloss for the Addis Ababa city realistic case scenario. | en_US |
| dc.identifier.uri | http://etd.aau.edu.et/handle/123456789/21116 | |
| dc.language.iso | en_US | en_US |
| dc.publisher | Addis Ababa University | en_US |
| dc.subject | Pathloss | en_US |
| dc.subject | FFBP | en_US |
| dc.subject | modeling | en_US |
| dc.subject | neural network | en_US |
| dc.subject | performance metrics | en_US |
| dc.subject | Propagation | en_US |
| dc.subject | Propagation Models | en_US |
| dc.title | Outdoor Propagation Pathloss Model for UMTS Networks using Feedforward Backpropagation Neural Network: Case of Addis Ababa, Ethiopia | en_US |
| dc.type | Thesis | en_US |