پديد آورنده :
اخباري، مهديه
عنوان :
رتبه بندي اعتباري مشتريان حقوقي بانك ها با رويكرد هوش مصنوعي
مقطع تحصيلي :
كارشناسي ارشد
محل تحصيل :
اصفهان: دانشگاه صنعتي اصفهان، دانشكده صنايع و سيستم ها
صفحه شمار :
يازده، 110، [II] ص.: مصور، جدول، نمودار
يادداشت :
ص. ع. به فارسي و انگليسي
استاد راهنما :
فريماه مخاطب رفيعي
توصيفگر ها :
ريسك اعتباري , صنعت بانكداري , مديريت ريسك , ANN
تاريخ نمايه سازي :
26/6/87
تاريخ ورود اطلاعات :
1396/03/06
رشته تحصيلي :
صنايع و سيستم ها
دانشكده :
مهندسي صنايع و سيستم ها
چكيده فارسي :
به فارسي و انگليسي: قابل رويت در نسخه ديجيتال
چكيده انگليسي :
ABSTRACT In the banking industry one issue that must be always considered by the credit policymakers is risk management Among various risks which banks are dealing with credit riskis most important It is caused by the losses of disability or lack of tendency of borrowersto pay their credit obligations To manage and control the mentioned risk credit rating systems are undeniablerequirement Such systems according to existent documents and information determinethe credit score of customers and rate them based on amount of their risk on bank It isevident that use of these systems helps bank to choose costumers in a good way Andthrough the control and reduction the credit risk improves efficiency level of providingbank facilities This study examines artificial intelligent based credit models consist of the artificialneural networks model adaptive Neuro fuzzy Inference Systems and a multi objectivefuzzy simplex genetic algorithm which is developed to optimize the fuzzy rules in fuzzyinference system are applied to predict bank legal customers financial performance After collecting and examining data 320 files related to legal customers of TEJARATbank branches in Tehran over 2001 2006 debt ratio operational ratio and return on equityratio were selected as explanatory variables And on the other side dependent variable wasconsidered as a dummy variable 0 for good credit and 1 for bad credit customers Thendata were divided in to model in sample and test out of sample sets After training anddeveloping models predictive performance of models is examined based on theirsensitivity and specificity ratios on the test set Empirical findings show that artificial neural network has highest accuracy atidentifying defaults in the portfolio out of sample Multi objective fuzzy simplex geneticalgorithm besides its good ability at identifying default non default cases has two moreadvantages First it is able to consider several objective functions in the training processand another is that its outcomes can be interpreted and find most effective explanatoryvariable on default Analysis was shown that debt ratio is the most consistent predictor ofdefault
استاد راهنما :
فريماه مخاطب رفيعي
