A Gaussian Process Regression Model to Predict Path Loss for an Urban Environment

This research paper presents a Gaussian process regression (GPR) model for predicting path loss signal in an urban environment. The Gaussian process regression model was developed using a dataset of path loss signal measurements acquired in two urban environments in Nigeria. Three different kernel functions were selected and compared for their performance in the Gaussian process regression model, including the squared exponential kernel, the Matern kernel, and the rotational quadratic kernel. The GPR model was validated and evaluated using various performance metrics and compared with different regression models. The results show that the Gaussian process regression model with the Matern kernel outperforms the linear regression and the support vector regression, but the decision tree and the random forest regression did better than the GPR in both cities. In the city of Port Harcourt, the GPR has a RMSE value of 3.0776 dB, the DTR has 2.0005 dB, the SVR has 3.6047 dB, the RFR has 1.0459 dB, and the LR 3.5947dB. The proposed GPR model provides more accurate and efficient approach to predict path loss compared to traditional methods. The extensive data collection and analysis conducted has resulted in a well-developed and accurate model.