Lighting fault prediction method based on LightGBM
XIN Guomao
GU Xueping
ZHU Yaru
LIU Dayang
HAO Jingquan
Abstract：This paper aims to address the inefficiency and lack of real-time capability in traditional lighting systems that rely on manual inspection,by proposing a lighting fault prediction method based on LightGBM.The study collects electrical parameters such as voltage,current,and power factor from streetlights,constructs a failure rate indicator,and extracts statistical features from historical data as input for modeling and prediction using the LightGBM algorithm.Experimental results show that the proposed method outperforms both linear regression and neural network models on the validation set,achieving a mean square error of 0.004 8 and a coefficient of determination of 0.969 7.The research demonstrates that this method can effectively predict the probability of streetlight failure with high accuracy and practicality,providing reliable support for predictive maintenance in smart lighting systems.
Keywords：fault predictionlamp source faultmachine learningLightGBM
Publication Date：2025-11-25
Pages：4( 77-80 )