Affinity and Class Probability based Fuzzy Support Vector Machine for Path Loss Prediction in 5G Communication System

Wireless propagation channel is a primary factor for optimizing and planning 5G communication system. Predicting path loss in 5G plays a vital role in optimizing performance of network and ensure appropriate communication by determining signals over distance. However, path loss remains a challenge in 5G system because of signal propagation in different environment which results in minimized and unreliable network performance. This research proposes Affinity and Class Probability based Fuzzy Support Vector Machine (ACPFSVM) to predict the path loss in 5G communication system. In the traditional SVM, the ACPF is involved for prediction to handle uncertainties efficiently in data which results in more robust and accurate performance. The Gaussian filter is applied to remove a noise by smoothening the images which minimize high-frequency noise while conserving significant structural features. Local Binary Pattern (LBP) and Gray Level Co-occurrence Matrix (GLCM) are applied to extract the features by capturing local texture patterns effectively. The proposed ACPFSVM obtains a less Mean Average Error (MAE) of 0.083 compared to existing methods like Random Forest (RF), Linear Regression (LR), and SVM respectively.