中輟生預測系統之探索-資料挖掘之應用; The Initial Research of The Predictive System of Dropouts - The Apply of Data-mining

NCUIR > School of Management at National Central University > Graduate Institute of Information Management > Electronic Thesis & Dissertation > Item 987654321/13012

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/13012

Title:	中輟生預測系統之探索-資料挖掘之應用;The Initial Research of The Predictive System of Dropouts - The Apply of Data-mining
Authors:	林志弘;Jyh-Horng Lin
Contributors:	資訊管理研究所
Keywords:	資料挖掘;中輟生預測;成本敏感預測;決策樹;非對稱資料分配;Dropouts Predict;Mining;Cost-sensitive Predict;Decision Tree
Date:	2003-05-23
Issue Date:	2009-09-22 15:21:36 (UTC+8)
Publisher:	國立中央大學圖書館
Abstract:	社會上常常傳出許多青少年的犯罪事件，根據研究顯示，涉案的青少年中，多屬「中輟生」，因此令人關切中輟問題的嚴重性。因為，中輟生的產生，涉及了社會與教育資源的浪費、以及中輟生的個人生涯發展等；而中輟生犯罪所帶給社會的傷害，更是難以估計。目前關於中輟生的研究大都是侷限於中輟成因的探討，以及中輟生的輔導及復學的相關研究，少有針對中輟行為預測的研究報告，所以本研究主要目的是探討資料挖掘應用於中輟生預測的可行性。本研究提出以成本敏感預測的分類觀念，利用問卷所收集到的實際學生中輟資料，以CART及C4.5兩種演算法進行實證研究，研究發現以實際資料分配比例預測，以CART演算法所建立的預測模型中，成本差異的提升可以有效提升中輟生預測的準確性，若是以對稱資料分配及專家投票決策方法進行預測時，成本差異提升並無法有效提升預測能力。 According to the researches, most of the dropouts are involved in the social criminal affairs. This situation are so concerned by us. This serious problem is something to with the waste of human power and educational resources. The hurt to our society of this problem is hard to count. At present, the researches of this problem just focus on the cause of forming, guidance and the career developing of personnel. Therefore, this research’s purpose is about the possibility of data-mining research of the predictive system of the dropouts. The approach of this research is to provide the predictive concept of the classification of cost-sensitive and use the questionnaire to collect the information of the real datum of the dropouts. Besides, This purpose of the research uses two different ways of algorithms, CART and C4.5, to approve its possibility so as to show the precise of the predictive system of dropouts could be effectively promote by the predictive model constructed by the CART algorithms, but “Unbalance Data Distribution Adjust Strategy” can not raise the predict effect of dropouts.
Appears in Collections:	[Graduate Institute of Information Management] Electronic Thesis & Dissertation

Files in This Item:

File	Size	Format

社群 sharing

Loading...