Development of Path Loss Prediction Model using Feature Selection-Machine Learning Approach

Abstract

Wireless network planning requires accurate coverage predictions to get good quality. The path loss accurate model requires a flexible model for each area including land and water. The purpose of this research is to develop a Cost-Hatta model that can be applied to the mixed land-water area. The approach used of this research is the three methods of feature selection of machine learning. The first stage of the research was the collection of field data. The measurement data included system, weather, and geographical parameters. The next stage was feature selection to obtain the best composition of features for the development of the model. The feature selection methods used were Univariate FS, Genetic Algorithm (GA), and Particle Swarm Optimization (PSO). After obtaining the best features from each method, the next stage was to form a model using four machine learning algorithms, namely Random Forest Regression (RF), Deep Neural Network (DNN), K-Nearest Neighbor Regression (KNN), and Support Vector Regression (SVR). The results of the improvements to the path loss prediction model were tested using the evaluation parameters of Root Mean Square Error (RMSE), Mean Square Error (MSE), and Mean Absolute Percentage Error (MAPE). The results of the testing showed that the improved Cost-Hatta model using the proposed Univariate-RF combination produced a very small RMSE value of 1.52. This indicates that the proposed model framework is highly suitable to be used in a mixed land-water area.