浙江省制造業(yè)注銷企業(yè)特征分析

信息計算科學論文編號:XXLW071 論文字數(shù):18247,頁數(shù):52

目  錄
中文摘要  i
英文摘要  ii
目錄 iii
第一章  引  言 1
第二章  業(yè)務分析 2
2.1  理論基礎 2
2.1.1  數(shù)據(jù)倉庫 2
2.1.2  商業(yè)智能 3
2.1.3  BusinessObjects Enterprise 3
2.2  數(shù)據(jù)條件 4
2.2.1  需求分析 4
2.2.2  對象選擇 4
第三章  數(shù)據(jù)處理 7
3.1  理論基礎 7
3.1.1  PL/SQL工具 7
3.1.2  數(shù)據(jù)挖掘 7
3.1.3  SPSS  Clementine 工具 8
3.2  數(shù)據(jù)準備 8
3.2.1  影響因素 8
3.2.2  數(shù)據(jù)選取 9
3.3  數(shù)據(jù)清洗 11
第四章  模型實現(xiàn) 14
4.1  理論基礎 14
4.1.1  決策樹概述 14
4.1.2  決策樹算法 14
4.1.3  決策樹優(yōu)勢 15
4.2  建立模型 15
4.2.1  C5.0模型 16
4.2.2  C&R Tree模型 17
第五章  模型分析 18
5.1  模型對比 18
5.2  C5.0模型分析 19
5.2.1  模型評價 19
5.2.2  結果分析 19
第六章  總結 24
6.1  論文主要工作總結 24
6.2  模型的改進 24
致  謝 25
參考文獻 26
附錄 27

摘  要
 在全球經(jīng)濟危機的背景下，洞察當今經(jīng)濟的潛在關系，對企業(yè)進行特征分析，可以清楚的了解企業(yè)當前狀況，是否面臨被注銷的風險，從而對有注銷特征的企業(yè)進行重點關注。
 本文從浙江省工商行政管理局數(shù)據(jù)庫中獲取企業(yè)數(shù)據(jù)，利用ETL（數(shù)據(jù)倉庫中的操作）對數(shù)據(jù)進行抽取、轉化以及裝載，從而得到年檢年份為2005至2007的浙江省企業(yè)數(shù)據(jù)。為進一步提高數(shù)據(jù)的有效性及真實性，文中利用SPSS Clementine 10.0工具，對數(shù)據(jù)進行進一步的篩選及清洗。本文基于數(shù)據(jù)挖掘理論，選擇C5.0 算法、C&R Tree兩種算法進行分析及比較，并結合浙江省企業(yè)數(shù)據(jù)的特征屬性，最終采用C5.0 算法建立的注銷企業(yè)特征分析模型。通過對模型求解及分析，結果表明浙江省制造業(yè)企業(yè)發(fā)生注銷主要由經(jīng)營年限，登記機關，企業(yè)規(guī)模等9個特征因素決定，其準確率達到95.81%，誤差率為4.29%。通過對誤差數(shù)據(jù)進行分析，得出模型產(chǎn)生誤差的主要原因是原始數(shù)據(jù)的缺失，這使得某些影響因素不能在模型中體現(xiàn)出來，使模型沒有達到最優(yōu)，但評估指標整體較好，對使用者能提供一定得支持。
 通過分析企業(yè)的特征，可以使決策者對企業(yè)的當前狀況了然于心，從而提出正確的，有利于各方面發(fā)展的信息。
 
關鍵詞：浙江工商；注銷企業(yè)；數(shù)據(jù)挖掘；C5.0 算法；C&R Tree

Abstract
 In the background of global economic crisis, Insight into the potential relationship between today's economy, analyze the features of enterprise, has a clear understanding of the current situation of the enterprise, Thereby determine whether the risk of being written off, and focus on the enterprise with the feature of written off.
 This article gain enterprise data from the database of Zhejiang Administration for Industry and Commerce, distill transform and load the data by ETL to gain zhejiang enterprise data between 2005 and 2007. In order to improve the Validity and authenticity of the data, filter the data using SPSS Clementine 10.0。The article is based on Data Mining Theory, analyze and compare C5.0 with C&R Tree algorithm , finally establish the feature analysis model of written_off enterprise using C5.0 algorithm . By answer and analyze the model,the results shows that  the appearance Manufacturing enterprises are written off depends on the operating life cycle, registration bodies and other nine factors. the accuracy rate up to 95.81%, the error rate up to 4.29%. By analyzing Error Data, we come to the result the main reason of error is lack of raw data,which makes some of the impact factor can not be reflected in the model, so that the model did not meet the best, but the overall eva luation are better, this may provide support to the users.
 By analyzing the characteristics of enterprises, the decision-makers can understand the current status clearer, thereby raise the right information which is benefit to all aspects of development.

Keywords：Zhejiang industrial and commercial, written_off enterprise, Data Mining, C5.0 algorithm, C&R Tree