Development of a Machine Learning Based Fault Detection Model for Received Signal Level in Telecommunication Enterprise Infrastructure

This research develops a machine-learning fault detection model for received signal levels in telecommunication infrastructure.The methodology involves modeling an enterprise point-to-multipoint wireless network using pathloss 5.0 software.Data from the simulated network, including free space pathloss, transmit power output, transmit antenna gain, transmitter loss, miscellaneous loss, and receiver loss, is used to train three regression models: gradient boosting regression (GBR), random forest regression (RFR), and K-Nearest Neighbor (KNN).The algorithm compares the received signal levels (RSL) of new data with a threshold value, triggering a "Fault" or "No-fault" condition.A "Fault" indicates a deviation in the RSL, prompting maintenance by the field support team.A "No-fault" means the RSL is within the accepted range, requiring no maintenance.Performance evaluation metrics such as mean absolute error (MAE), mean square error (MSE), R-squared, and root mean square error (RMSE) were compared to select the optimal model.Experimental results show that the RFR model outperforms GBR and KNN with MAE: 0.007101, MSE: 0.000610, R-squared: 0.999992, and RMSE: 0.024697.Leveraging these machine learning-based fault detection models enables telecom service providers to optimize network performance, reduce downtime, and increase customer satisfaction.