西南地区以其庞大的人口基数和经济规模著称,然而经济结构转型引发的结构性失业问题日益凸显。失业问题不仅加重了民众经济负担,还影响了社会稳定和劳动力资源的有效配置。失业率作为评估区域经济健康的关键指标,对社会稳定、民众生活...西南地区以其庞大的人口基数和经济规模著称,然而经济结构转型引发的结构性失业问题日益凸显。失业问题不仅加重了民众经济负担,还影响了社会稳定和劳动力资源的有效配置。失业率作为评估区域经济健康的关键指标,对社会稳定、民众生活质量和经济政策规划具有直接影响。为科学预测区域失业率,本文基于1997年至2023年西南地区的面板数据,构建了Gamma回归边际模型,并选用对数函数作为连接函数,使用R软件采用了AR (1)自相关、独立相关以及可交换相关工作矩阵结构下的广义估计方程(GEE)以及惩罚广义估计方程(PGEE)这6种方法进行了实证分析。结果显示,当惩罚参数λ设定为0.13时,采用PGEE-AR (1)自相关结构方法构建的模型预测性能最好,它的测试集MSE、MAE、MAPE分别为0.207、0.32、8.85,都小于其它5中方法。这一研究为经济转型期的失业率预测提供了科学依据。Southwest China, renowned for its large population and economic scale, faces increasingly prominent structural unemployment issues triggered by economic restructuring. Unemployment not only exacerbates financial burdens on residents but also threatens social stability and the efficient allocation of labor resources. As a critical indicator of regional economic health, the unemployment rate directly impacts social stability, quality of life, and economic policy planning. To scientifically predict regional unemployment trends, this study utilizes panel data from Southwest China (1997~2023) to construct a Gamma regression marginal model with a logarithmic link function. Employing R software, six methodological approaches—Generalized Estimating Equations (GEE) and Penalized Generalized Estimating Equations (PGEE) under AR (1) autocorrelation, independent, and exchangeable working correlation matrix structures—were applied for empirical analysis. Results show that the model constructed using the PGEE-AR (1) method with autocorrelation structure, with a penalty parameter λset to 0.13, achieved the best prediction performance. Its test set MSE, MAE, and MAPE are 0.207, 0.32, and 8.85%, respectively, all lower than those of the other five methods. This research provides a scientific basis for unemployment rate prediction during the economic transition period.展开更多
文摘西南地区以其庞大的人口基数和经济规模著称,然而经济结构转型引发的结构性失业问题日益凸显。失业问题不仅加重了民众经济负担,还影响了社会稳定和劳动力资源的有效配置。失业率作为评估区域经济健康的关键指标,对社会稳定、民众生活质量和经济政策规划具有直接影响。为科学预测区域失业率,本文基于1997年至2023年西南地区的面板数据,构建了Gamma回归边际模型,并选用对数函数作为连接函数,使用R软件采用了AR (1)自相关、独立相关以及可交换相关工作矩阵结构下的广义估计方程(GEE)以及惩罚广义估计方程(PGEE)这6种方法进行了实证分析。结果显示,当惩罚参数λ设定为0.13时,采用PGEE-AR (1)自相关结构方法构建的模型预测性能最好,它的测试集MSE、MAE、MAPE分别为0.207、0.32、8.85,都小于其它5中方法。这一研究为经济转型期的失业率预测提供了科学依据。Southwest China, renowned for its large population and economic scale, faces increasingly prominent structural unemployment issues triggered by economic restructuring. Unemployment not only exacerbates financial burdens on residents but also threatens social stability and the efficient allocation of labor resources. As a critical indicator of regional economic health, the unemployment rate directly impacts social stability, quality of life, and economic policy planning. To scientifically predict regional unemployment trends, this study utilizes panel data from Southwest China (1997~2023) to construct a Gamma regression marginal model with a logarithmic link function. Employing R software, six methodological approaches—Generalized Estimating Equations (GEE) and Penalized Generalized Estimating Equations (PGEE) under AR (1) autocorrelation, independent, and exchangeable working correlation matrix structures—were applied for empirical analysis. Results show that the model constructed using the PGEE-AR (1) method with autocorrelation structure, with a penalty parameter λset to 0.13, achieved the best prediction performance. Its test set MSE, MAE, and MAPE are 0.207, 0.32, and 8.85%, respectively, all lower than those of the other five methods. This research provides a scientific basis for unemployment rate prediction during the economic transition period.
