【作者】 罗艳；

【导师】 陈丹；

【作者基本信息】 云南大学 ， 统计学， 2021， 硕士

【摘要】 随着我国经济的不断发展,人民生活水平不断提高,城市生活垃圾产量不断增长,“垃圾围城”现象不断涌现,城市环境卫生、居民健康受到严重威胁。城市生活垃圾产量是城市管理者处理生活垃圾相关问题重要的参考数据,因此对生活垃圾产量预测的需求不断增长,要求不断提高。目前国内外关于城市生活垃圾产量预测的方法很多,主要有多元线性回归、灰色预测和时间序列方法,它们通常是根据选定的单一模型进行预测。单一模型的预测,常常忽视了模型不确定性。本文选用模型平均方法研究云南省城市生活垃圾产量,充分考虑了模型的不确定性。本文从云南省城市生活垃圾产量影响因素入手,选择10个具有代表性的解释变量,用生活垃圾清运量代表生活垃圾产量作为被解释变量,共拟合1024个线性模型,然后分别用频率模型平均和贝叶斯模型平均方法对云南省的城市生活垃圾产量进行预测,并以相对误差、最优率和相对误差标准差作为评价指标,分析比较各种方法的优劣,再选用贵州省城市生活垃圾相关数据对模型进行稳健性研究,最后选取最优权重模型平均方法对云南省2021-2025年的生活垃圾产量进行预测。结果表明:(1)对于5种频率模型平均方法,最优权重模型平均法无论从相对误差、最优率还是相对误差标准差考虑,均表现最优;综合考虑三种指标表现次优的是Jackknife模型平均方法和Smoothed AIC方法,相对较差的是Mallows模型平均方法和Smoothed AIC方法。(2)对于贝叶斯模型平均方法,不同模型先验和参数先验下,国内生产总值、餐饮企业数、居民消费价格指数、商品零售价格指数仍保持很强的解释能力。(3)两类模型平均方法比较,最优权重模型平均法仍然保持最优,贝叶斯模型平均方法在所有方法中属于中间水平。(4)模型平均方法对生活垃圾产量预测的稳健性较好,改变数据集预测结论保持一致。(5)云南省2021-2025年生活垃圾产量预测值分别为536.39万吨、571.20万吨、608.49万吨、648.35万吨、690.84万吨,呈逐年递增的趋势。更多还原

【Abstract】 With the continuous development of China’s economy,people’s living standards continue to improve,municipal solid waste(MSW)output continues to grow,"garbage siege" phenomenon continues to emerge,and urban environmental hygiene and residents’ health are seriously threatened.The output of MSW is an important reference data for city managers to deal with the problems related to MSW.Therefore,the demand for predicting the output of MSW is constantly increasing and the requirements are constantly improved.At present,there are many domestic and foreign research about MSW output forecasting methods,including multiple linear regression,grey forecasting and time series methods,which are usually based on the selected single model.Uncertainty of the model is often ignored in the prediction of a single model.In this paper,the model averaging method was used to study the MSW output in Yunnan province,which fully considered the uncertainty of the model.Starting with the influencing factors of MSW output in Yunnan province,this paper selects 10 representative explanatory variables,and the volume of MSW collection represented the output of MSW as the explained variable,and fitting a total of 1024 linear models,and then respectively with frequency model averaging method and bayesian model averaging method to forecast the output of MSW in Yunnan province,and the relative error,the optimal rate,relative standard deviation of error as the evaluation index,analysis of comparative advantages and disadvantages of various methods.Then use the relevant data of MSW in Guizhou Province to study the robustness of the model,and finally select the optimal weight model averaging method to predict the output of MSW in Yunnan Province from 2021 to 2025.The results show that :(1)For the five frequency model averaging methods,the optimal weighted model averaging method performs best in terms of relative error,optimal rate and standard deviation of relative error;Taking the three indicators into account,the Jackknife model averaging method and the Smoothed AIC method show the inferior performance,while the Mallows model averaging method and the Smoothed AIC method are relatively poor.(2)For Bayesian model averaging method,under different model priors and parameter priors,GDP,number of catering enterprises,consumer price index and retail price index still maintain strong explanatory ability.(3)Compared with the two kinds of model averaging methods,the optimal weighted model averaging method is still the best,and the Bayesian model averaging method is at the middle level among all the methods.(4)The model averaging method has good robustness in the forecast of MSW output,and the prediction conclusions of changing data sets are consistent.(5)The forecasted value of MSW output in Yunnan Province from 2021 to2025 is 5.3639 million tons,5.712 million tons,6.0849 million tons,6.483 million tons,and 6.908 million tons respectively,showing an increasing trend year by year.更多还原