پديد آورنده :
رحماني، اصغر
عنوان :
شناسايي نقاط پرت در ابعاد بزرگ
مقطع تحصيلي :
كارشناسي ارشد
گرايش تحصيلي :
آمار رياضي
محل تحصيل :
اصفهان: دانشگاه صنعتي اصفهان، دانشكده علوم رياضي
صفحه شمار :
يازده،104ص.: جدول،نمودار
يادداشت :
ص.ع.به فارسي و انگليسي
استاد راهنما :
سروش عليمرادي
استاد مشاور :
علي زينل همداني
توصيفگر ها :
برآوردگرهاي استوار , مولفه هاي اصلي استوار
تاريخ نمايه سازي :
8/11/92
استاد داور :
علي رجالي، هوشنگ طالبي
چكيده فارسي :
به فارسي و انگليسي: قابل رويت در نسخه ديجيتالي
چكيده انگليسي :
Outlier identi cation in high dimensions Asghar Rahmani a rahmani@math iut ac ir 2013 Department of Mathematical Sciences Isfahan University of Technology Isfahan 84156 83111 Iran Supervisor Dr Soroush Alimoradi salimora@cc iut ac ir Advisor Dr Ali Zeinal Hamadani hamadani@cc iut ac ir 2010 MSC 62J20 62F35 Keywords Outlier identi cation Robust estimators High dimension Robust principal compo nents AbstractOutliers are data points that lying far away from the main part of a data set and probably not follow ing the assumed model These data points are often the special points of interest in many practicalsituations and their identi cation is the main purpose of the investigation Moreover accurate iden ti cation of outliers plays an important role in statistical analysis since they can strongly in uencethe classical methods and even falsify results Outliers have di rent sources they may be the resultof an error arise from the inherent variability of the data set i e extrem values from the tails of thedistribution or may be generated from another model Outliers are found frequently in real data andare more likely to occur in data sets with many observations and or variables and often they do notshow by simple visual inspection However there are two basic approaches to outlier identi cation inmultivariate observations These methods are distance based and projection pursuit Distance basedmethods aim to detect outliers by computing a measure of how far a particular point is from the centerof the data The usual measure of outlyingness for a data point xi Rp i 1 n is Mahalanobisdistance xi x S 1 xi x M Di where x and S are arithmetic mean and sample covariance matrix Classical tools based on the mean and covariance matrix are rarely able to detect all the multivariate outliers in a given sample dueto the masking e ect To avoid this e ect the goal of robust methods is to develop approaches that
استاد راهنما :
سروش عليمرادي
استاد مشاور :
علي زينل همداني
استاد داور :
علي رجالي، هوشنگ طالبي
