Light Gradient Boosting Machine-Based Link Quality Prediction for Wireless Sensor Networks

Abstract

Link quality prediction is a fundamental component of the wireless network protocols and is essential for routing protocols in wireless sensor networks (WSNs). Effective link quality prediction can select high-quality links for communication and improve the reliability of data transmission. In order to improve the accuracy of the link quality prediction model and reduce the model complexity, the link quality prediction model based on the light gradient boosting machine (LightGBM-LQP) is proposed in this paper. Specifically, agglomerative hierarchical clustering and manual division are combined to grade the link quality and obtain the labels of samples. Then, light gradient boosting machine (LightGBM) classification algorithm and Focal Loss are used to estimate the link quality grades. In order to reduce the impact of data imbalance, Borderline-SMOTE is employed to oversample the minority link quality samples. Finally, LightGBM-LQP predicts link quality grade at the next moment with historical link quality information. The experimental results on data collected from a real-world WSNs show that the proposed model has better prediction accuracy and shorter predicting time compared to related models.

Advanced Machine Learning Techniques for Accurate Very-Short-Term Wind Power Forecasting in Wind Energy Systems Using Historical Data Analysis

“…The library known as LightGBM, which stands for "Light Gradient Boosted Machine", was developed at Microsoft and offers a productive implementation of the gradient boosting technique. The most significant advantage provided by the LightGBM is the modification of the training algorithm, which not only makes the process considerably quicker but also, in many instances, produces a more accurate model [39]. The LightGBM method takes as input a supervised training set X and a loss function L(y, f (x)) whose anticipated value is to be minimized f (x).…”

Algorithm for link prediction in self-regulating network with adaptive topology based on graph theory and machine learning

“…Применительно к КФС и сенсорным сетям, алгоритмы предсказания связей часто используются для прогнозирования маршрутов в сетях, предсказания сбоев и решения подобных проблем, связанных с надежностью. Так, авторы работы [23] также используют link prediction для прогнозирования качества каналов связи как одной из важнейших задач маршрутизации в сложных сетях.…”