Share
مقارنة بين الخوارزمية الجينية والشبكات العصبية في تقدير موقع الوسيط لنماذج الانحدار متعدد المتغيرات اللامعلمي == Compared Between Genetic Algorithm And Neural Networks To Estimate The Model of Multivariate Nonparametric
Author name:
فاطمة عبد الحميد جواد البيرماني
Supervisor name:
صباح منفي رضا الشمري
General topic:
Administration and Economics
Specific topic:
Statistics
Degree:
Doctorate
University:
University of Baghdad - Faculty Of Administration And Economics - Department Of Statistics
Language:
Arabic
University location:
Baghdad
First pages:
07T3934 - p.pdf
Abstract:
تستعمل الطرق اللامعلمية في البيانات التي تحتوي على قيم شاذة، الاهمية الاساسية في استعمال الطرق اللامعلمية هو تحديد موقع الوسيط، ففي انموذج انحدارمتعدد المتغيرات يكون من الصعوبة تحديد موقع الوسيط لوجود اكثر من بعد وتشتت القيم وزيادة بيانات الظاهرة المدرو | Used Nonparametric methods to estimates the data containing outlier values ,where classical statistical methods are affected in estimation by having these values,the fundamental importance in the use of Nonparametric methods in the estimate is to identify the median location ,in the multivariate model be difficult to identify the location ,because the model contain more than distance , dispersion of the values and increase sample size.In this thesis it has been the application of genetic algorithms and neural network to find estimate for the median location so dependent on a minimum covariance determinant as one of nonparametric method robust in estimate.The comparison was made between genetic algorithm and neural network in determining the best way to give more accurate results and faster as especially when increasing the size of the sample in addition to the proposed methods and most important characteristics of genetic algorithms and neural networks , possibility of merging with each other to generate a new generation of genetic algorithms and neural networks , after determining the best way of comparison between the genetic algorithm and nural network as well as between the proposed roads are estimated coefficients of the model to the changing time - smoothing spline using either paved parameters have been estimated in manner CV. The study has been applied to environmental pollution statistics to drinking water for the year (2013) included all of Iraqs’ provinces except for the Kurdistan region , divided into (10) months have been used (9) the types of chemical indicators and physical causing contamination of drinking water at the minimum of the measure exceeded.The most important conclusions , the application of genetic algorithm Fast - MCD - Nested Extension where better than MWCDgenetic algorithm , either the proposed genetic algorithm resulting from the merger between Back propagation and above results were better than MWCD genetic algorithm , in terms of neural networks , the use of neural network ART were more the accuracy and speed of multilayered neural network Back propagation and from genetic algorithm Fast - Nested Extension , either neural network resulting from merger between the network ART , Multilayered neural network as well as compared with the proposed genetic algorithm.
