Mobility Prediction Algorithms for Handover Management in Heterogeneous LiFi and RF Networks: An Ensemble Approach

Abstract

Light Fidelity (LiFi) is a communication technology that operates in the Visible Light (VL) region, using light as a medium to enable ultra-high-speed communication. The spectrum occupied by LiFi does not overlap with the Radio Frequency (RF) spectrum. Thus, they can be used in a hybrid manner to enhance the Quality of Service (QoS) for users. However, in a heterogeneous LiFi and RF network, users experience constant handovers due to the small coverage area of the LiFi and their frequent movement. This study proposes an intelligent handover scheme, where the network parameters of the users are used to train four machine learning models, namely an Artificial Neural Network (ANN), an Adaptive Neurofuzzy Inference System (ANFIS), a Support Vector Machine (SVM), and a Regression Tree (RT), to predict the mobility of the users, so that the central network can have a priori mobility information to ensure seamless connectivity. Furthermore, the performance of the standalone models was enhanced by integrating ensemble learning techniques such as the Simple Averaging Ensemble (SAE), Weighted Averaging Ensemble (WAE), and a Meta-Learning Ensemble (MLE). The results show that the ensemble algorithms improved prediction performance, with an average error decrease of 44.40%, 53.53%, and 61.03% for SAE, WAE, and MLE, respectively, which further demonstrated the effectiveness and robustness of using ensemble algorithms to predict user mobility.