控制理论与应用
控製理論與應用
공제이론여응용
CONTROL THEORY & APPLICATIONS
2009年
12期
1365-1370
,共6页
粗糙集%支持向量机%神经网络%违约判别
粗糙集%支持嚮量機%神經網絡%違約判彆
조조집%지지향량궤%신경망락%위약판별
rough set%support-vector-machines%neural network%default prediction
建立了粗糙集和支持向量机集成的企业贷款违约判别模型,该模型首先利用自组织映射 (SOM)神经网络对具有连续属性值的财务数据进行离散处理,并应用遗传算法约简评价指标,然后将约简得到的最小条件属性集及相应的原始数据送入支持向量机进行训练,最后对企业短期贷款检验样本进行违约判别.采用贷款企业数据库558家制造业样本企业和522家房地产业样本企业进行交叉验证的实证研究,结果表明,与BP神经网络、多元判别分析、Logistic等违约判别模型相比,粗糙集和支持向量机集成的违约判别模型有更好的预测效果.
建立瞭粗糙集和支持嚮量機集成的企業貸款違約判彆模型,該模型首先利用自組織映射 (SOM)神經網絡對具有連續屬性值的財務數據進行離散處理,併應用遺傳算法約簡評價指標,然後將約簡得到的最小條件屬性集及相應的原始數據送入支持嚮量機進行訓練,最後對企業短期貸款檢驗樣本進行違約判彆.採用貸款企業數據庫558傢製造業樣本企業和522傢房地產業樣本企業進行交扠驗證的實證研究,結果錶明,與BP神經網絡、多元判彆分析、Logistic等違約判彆模型相比,粗糙集和支持嚮量機集成的違約判彆模型有更好的預測效果.
건립료조조집화지지향량궤집성적기업대관위약판별모형,해모형수선이용자조직영사 (SOM)신경망락대구유련속속성치적재무수거진행리산처리,병응용유전산법약간평개지표,연후장약간득도적최소조건속성집급상응적원시수거송입지지향량궤진행훈련,최후대기업단기대관검험양본진행위약판별.채용대관기업수거고558가제조업양본기업화522가방지산업양본기업진행교차험증적실증연구,결과표명,여BP신경망락、다원판별분석、Logistic등위약판별모형상비,조조집화지지향량궤집성적위약판별모형유경호적예측효과.
An integrated model of rough sets and support-vector-machines for the default prediction of short-term loan is proposed. The financial data is discretized by using self-organizing mapping neural network; and the evaluation indices are reduced with no information loss through genetic algorithm. The reduced indices together with relevant data are used to train support-vector-machines and discriminate between healthy and default testing samples. 558 manufacturing industry's loan firms and 522 real estate industry's loan firms are selected as test samples, The prediction accuracy of the integrated model combining rough sets and support-vector-machines is better than that of other methods such as BP neural network, multiple discriminant analysis and logistic regression.
