DEVELOPMENT OF A MACHINE LEARNING MODEL FOR CLASSIFYING FREE SPACE OPTICS CHANNEL IMPAIRMENTS

BABATUNDE, KAREEM SUNDAY (2022) DEVELOPMENT OF A MACHINE LEARNING MODEL FOR CLASSIFYING FREE SPACE OPTICS CHANNEL IMPAIRMENTS. Masters thesis, Landmark University, Omu Aran, Kwara State.

Text Babatunde Kareem Research Thesis 020822 CORRECTED.docx Download (5MB)

Abstract

Free Space Optics is an optical communication method that uses Free Space instead of Fibre Cable to convey data through a medium from a transmitter to the receiver. It is a viable solution for ensuring high data rates and last-mile communication delivery in Next-Generation wireless communication. However, adverse weather conditions can significantly impair the performance of FSO channel links during transmission. Recently, Machine Learning models have received lots of attention in proffering solutions to signal impairments (that is, atmospheric turbulence, noise, and pointing errors) in optical networks. Machine Learning is a subset of Artificial Intelligence that deals with extracting patterns from data and then using those patterns to enable algorithms to improve on the experience. It allows computers to learn without having to be explicitly programmed. In this work, the K-Means clustering algorithm combined with Support Vector Machine (SVM) and K Nearest Neighbour (KNN) classifiers were used, trained, and tested for classifying the channel impairments in FSO links. The Dataset used for the training and testing of the models is fetched from an open-source called “Kaggle”, cleaned by applying pre-processing techniques, and transformed before being used in the model via MATLAB simulation. Binary Classification Evaluation metrics, such as Accuracy, Precision, Specificity, Sensitivity, and F1 score were used in conjunction with the Confusion Matrix to determine the values of True Positive (TP), True Negative (TN), Faise Negative (FN), False Positive (FP) in calculating the expression of the performance evaluation metrics. The Performance metrics comparison between the two classifiers (K-Means/SVM and K-Means/KNN) suggests that K-means/SVM outperformed K Means/KNN with 99.2% accuracy. The preferred model (K-Means/SVM) is also seen to outperform some existing classification models (K-means with Fuzzy Logic and Random Forest) during the comparison The research work developed a Machine Learning model for the classification of Free Space Optical Impairments such as atmospheric turbulence, noise and pointing error to the accuracy of 99.2% and also provides an effective tools for Free Space Optical (FSO) equipment manufacturers and for the effective monitoring and mitigation of losses of transmitted information in the communication industry.

Actions (login required)

View Item