學術產出-研究報告
文章檢視/開啟
書目匯出
-
題名 隨機模型與電腦模擬的人口推估研究 作者 余清祥;蔡紋琦 貢獻者 行政院經濟建設委員會
國立政治大學統計學系關鍵詞 人口推估;人口變動要素合成方法;隨機推估;電腦模擬;專家意見;交叉驗證;拔靴法;預測;Population Projection;Cohort Component;Stochastic Projection;Computer Simulation;Expert Opinion;Cross Validation;Bootstrap;Forecast;統計學;社會學;經濟學 日期 2008 上傳時間 11-十一月-2010 13:34:40 (UTC+8) 摘要 人口推估涉及國家的政策及規劃，精確的結果可協助國家適時制訂政策，提高國民福祉。臺灣現在使用的方法為人口變動要素合成方法(Cohort Component Method)，可算是情境推估的一種，其來源追溯至1930年代，一般分成高、中、低推計三種可能，其中高、低推計代表推估數值的可能範圍，中推計為最有可能的未來結果。但這種方法過於依賴不具有機率意涵的主觀專家意見，近年來聯合國、歐美各國引入隨機方法，修正情境推估的缺點。 修正情境推估的方法大致可分為以下三種類型：一為隨機推估(Stochastic Forecast)、一為模擬情境(Random Scenario)、一為推估誤差(ex post Method)。這三種方法各有其優勢，尚無絕對優劣的區別，但可以確定的是他們都優於傳統的情境推估。本計畫將整理這三種新的推估方法，並建立相關的理論模型，引進這三種方法於國內的人口推估研究。 除了整理新的推估方法，本計畫也將使用臺灣地區、其他國家的資料，以實證分析、藉由交叉驗證比較三種方法的特色；另外，也將以電腦模擬，仿造資料採礦中常用的「估計-測試」(Training and Testing)的判斷準則，評估方法的優劣。這些結果也將與傳統的情境推估比較，確定新推估方法確實較佳。本研究最後的結果除了提供學術界參考，也將詳細解說研究心得及注意事項，提供臺灣地區人口推估之用。 Population projections are essential to government planning and policy making. They can help to determine whether or not the policies match the needs of citizens and the planning needs any further modifications. The current projection method used in Taiwan is Cohort Component method and the assumption involved can be treated as scenario projection. Scenario projection was first proposed in the early 1930s. In scenario projection, the possible trend of future populations is usually divided into high, medium, and low projections. The high and low projections cover the possible ranges of future populations, and the medium projection is the most likely possibility. The problem is that, the scenario projection often relies solely on expert opinions. As a result, this projection usually can not be interpreted in the sense of probability. In recent years, most countries, such as the U.S. and the UN, have modified the scenario projection in order to provide the result of population projection with a broader prospect. There are three ways of modifications: Stochastic Forecast, Random Scenario, and ex post Methods. These methods have broader prospects than the traditional scenario projection and can be used together to improve the quality of cohort component method. In this project, we will first review and summarize these methods. Then, based on the results, we will propose related models for population projection in Taiwan. In addition to summarizing methods, we will use data from Taiwan and other countries for empirical studies, to evaluate the performances of the modifications via cross-validation. We will also use computer simulation to evaluate the performances, based on the criterion of Training-Testing in the field of data mining. All these results will be compared to those of traditional population projection method, and be used to verify if the modifications have better performances. The final outcomes of this project will be published in public and the techniques will be transferred to the government for the use of future population projection. 關聯 CEPD 96032
基礎研究
委託研究
研究期間: 9606~9706
研究經費: 1423 千元資料類型 report dc.contributor 行政院經濟建設委員會 en_US dc.contributor 國立政治大學統計學系 en_US dc.creator (作者) 余清祥;蔡紋琦 zh_TW dc.date (日期) 2008 en_US dc.date.accessioned 11-十一月-2010 13:34:40 (UTC+8) - dc.date.available 11-十一月-2010 13:34:40 (UTC+8) - dc.date.issued (上傳時間) 11-十一月-2010 13:34:40 (UTC+8) - dc.identifier.uri (URI) http://nccur.lib.nccu.edu.tw/handle/140.119/47990 - dc.description.abstract (摘要) 人口推估涉及國家的政策及規劃，精確的結果可協助國家適時制訂政策，提高國民福祉。臺灣現在使用的方法為人口變動要素合成方法(Cohort Component Method)，可算是情境推估的一種，其來源追溯至1930年代，一般分成高、中、低推計三種可能，其中高、低推計代表推估數值的可能範圍，中推計為最有可能的未來結果。但這種方法過於依賴不具有機率意涵的主觀專家意見，近年來聯合國、歐美各國引入隨機方法，修正情境推估的缺點。 修正情境推估的方法大致可分為以下三種類型：一為隨機推估(Stochastic Forecast)、一為模擬情境(Random Scenario)、一為推估誤差(ex post Method)。這三種方法各有其優勢，尚無絕對優劣的區別，但可以確定的是他們都優於傳統的情境推估。本計畫將整理這三種新的推估方法，並建立相關的理論模型，引進這三種方法於國內的人口推估研究。 除了整理新的推估方法，本計畫也將使用臺灣地區、其他國家的資料，以實證分析、藉由交叉驗證比較三種方法的特色；另外，也將以電腦模擬，仿造資料採礦中常用的「估計-測試」(Training and Testing)的判斷準則，評估方法的優劣。這些結果也將與傳統的情境推估比較，確定新推估方法確實較佳。本研究最後的結果除了提供學術界參考，也將詳細解說研究心得及注意事項，提供臺灣地區人口推估之用。 Population projections are essential to government planning and policy making. They can help to determine whether or not the policies match the needs of citizens and the planning needs any further modifications. The current projection method used in Taiwan is Cohort Component method and the assumption involved can be treated as scenario projection. Scenario projection was first proposed in the early 1930s. In scenario projection, the possible trend of future populations is usually divided into high, medium, and low projections. The high and low projections cover the possible ranges of future populations, and the medium projection is the most likely possibility. The problem is that, the scenario projection often relies solely on expert opinions. As a result, this projection usually can not be interpreted in the sense of probability. In recent years, most countries, such as the U.S. and the UN, have modified the scenario projection in order to provide the result of population projection with a broader prospect. There are three ways of modifications: Stochastic Forecast, Random Scenario, and ex post Methods. These methods have broader prospects than the traditional scenario projection and can be used together to improve the quality of cohort component method. In this project, we will first review and summarize these methods. Then, based on the results, we will propose related models for population projection in Taiwan. In addition to summarizing methods, we will use data from Taiwan and other countries for empirical studies, to evaluate the performances of the modifications via cross-validation. We will also use computer simulation to evaluate the performances, based on the criterion of Training-Testing in the field of data mining. All these results will be compared to those of traditional population projection method, and be used to verify if the modifications have better performances. The final outcomes of this project will be published in public and the techniques will be transferred to the government for the use of future population projection. en_US dc.format application/pdf en_US dc.format.extent 5345762 bytes - dc.format.mimetype application/pdf - dc.language zh-TW en_US dc.language.iso en_US - dc.relation (關聯) CEPD 96032 en_US dc.relation (關聯) 基礎研究 en_US dc.relation (關聯) 委託研究 en_US dc.relation (關聯) 研究期間: 9606~9706 en_US dc.relation (關聯) 研究經費: 1423 千元 en_US dc.subject (關鍵詞) 人口推估;人口變動要素合成方法;隨機推估;電腦模擬;專家意見;交叉驗證;拔靴法;預測;Population Projection;Cohort Component;Stochastic Projection;Computer Simulation;Expert Opinion;Cross Validation;Bootstrap;Forecast;統計學;社會學;經濟學 en_US dc.title (題名) 隨機模型與電腦模擬的人口推估研究 zh_TW dc.type (資料類型) report en
