SARIMA与Prophet的混合算法在时间序列预测中的应用研究  

A Hybrid Model Based on SARIMA and Prophet for Improving Time Series Forecasting Accuracy

作　　者：李长生 LI Changsheng(College of Information Engineering,Lanzhou Petrochemical Vocational and Technical University,Lanzhou Gansu 730060)

机构地区：[1]兰州石化职业技术大学信息工程学院,甘肃兰州730060

出　　处：《软件》2025年第1期7-9,共3页Software

基　　金：兰州石化职业技术大学科研项目“时间序列模型在黄河水沙环境监测中的应用研究”(2023KY-28);兰州石化职业技术大学科研项目“机器学习技术在中药材鉴别分析中的应用研究”(2024KY-19)阶段性成果之一。

摘　　要：本研究提出了一种基于SARIMA与Prophet模型的混合算法,以提升时间序列预测的精度。SARIMA模型擅长处理线性趋势和季节性变化,Prophet模型则适用于捕捉非线性趋势和异常波动。将两者结合后的混合算法能够更全面地对复杂数据进行预测。在零售、气象和金融市场等行业中的实验结果表明,混合算法在预测准确性上优于单一模型,误差率降低了15%以上。本文进一步探讨了混合算法的设计、优化和实际应用,为未来时间序列预测提供了新的思路和方法。This study proposes a hybrid algorithm combining SARIMA and Prophet models to enhance time series forecasting accuracy.The algorithm leverages SARIMA's ability to handle seasonality and linear trends,while Prophet captures nonlinear trends,overcoming the limitations of individual models.The hybrid algorithm that combines the two can predict complex data more comprehensively.Experimental results show that the hybrid model significantly outperforms SARIMA and Prophet alone,particularly in datasets with complex trends and seasonal fluctuations,reducing prediction errors.Furthermore,the hybrid algorithm is applicable not only in industries with clear seasonality,such as retail and meteorology,but also in financial markets where volatility and anomalies are prevalent.Real-world case studies validate the model's broad applicability and flexibility.

关 键 词：时间序列预测 SARIMA模型 Prophet模型 混合算法 预测准确性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]