摘要
为具有多个季节周期的时间序列生成预测是当今许多行业的一个重要用例。在这些情况下,要产生更准确和有意义的预测,就必须考虑多季节模式。在本文中,我们提出了一种基于多季节性分解的预测框架——季节性梯度提升决策树(P-GBDT)来预测具有多个季节模式的时间序列。我们的方法通过结合多季节分解技术,以补充GBDT的学习过程。实验表明,季节性分解步骤是有益的,可以在准确性和偏差方面提供更好的预测。
Making predictions for time series with multiple seasonal cycles is an important use case in many industries today.In these cases,multi-seasonal models must be considered to produce more accurate and meaningful predictions.In this paper,we propose a prediction model-Seasonal Gradient Boosted Decision Tree(P-GBDT)-based on multi-seasonal decomposition to predict time series with multiple seasonal patterns.Our approach complements the GBDT learning process by combining multi-seasonal decomposition techniques.It shows that the procedures of seasonal decomposition are beneficial to the better predictions in terms of accuracy and errors.
作者
钟祖建
ZHONG Zu-jian(Central China Normal University,Wuhan Hubei 430000,China)
出处
《新一代信息技术》
2020年第13期7-11,共5页
New Generation of Information Technology
