基于两阶段多变点估计对动力煤价格的实证分析
Empirical Analysis of Thermal Coal Price Based on Two-Stage Multiple Change-Point Estimation

将变量选择技术应用到变点估计中可以同时得到多变点数量和位置的估计结果，大大降低了计算复杂度。文章首先介绍了两阶段多变点检测与估计方法的基本概念，以2021年12月22日至2024年11月6日的秦皇岛动力煤价格作为研究对象，并对其中的训练集建立线性回归模型。其次，利用此变点估计方法估计线性回归模型中系数的多变点，根据变点个数和位置的估计结果建立更精准的分段线性回归模型，并运用最后一个子段的线性回归模型对测试集进行短期预测。实证结果表明，分段线性回归模型能更精准地刻画和预测秦皇岛动力煤价格的走势。
Applying variable selection techniques to multiple change-point estimation can yield estimates of the number and location of multiple change-point at the same time, significantly reducing computational complexity. The article first introduces the basic concepts of two-stage multiple change-point detection and estimation methods, takes Qinhuangdao thermal coal prices from December 22, 2021 to November 6, 2024 as the research object, and builds a linear regression model for the training set therein. Secondly, this change-point estimation method is utilized to estimate the multiple change-point of the coefficients in the linear regression model, establish more accurate segmented linear regression modeling based on the estimation of the number and location of change-point, and apply the linear regression model of the last subsection to make short-term predictions about the test set. The empirical results show that the segmented linear regression model can more accurately portray and predict the trend of Qinhuangdao thermal coal prices.