مقارنة بين الخوارزمية الجينية MWCD و الشبكة العصبية متعددة الطبقات الارجاعية (الارتدادية)‎ لتحديد المؤشرات المسببة لتلوث المياه

Abstract EN

Nonparametric methods are used in the data that contain outliers values.

The main importance in using Nonparametric methods is to locate the median in the multivariate regression model.

It is difficult to locate the median due to the presence of more than one dimension and the dispersion of values and the increase of the studied phenomenon data.

The genetic algorithms Minimum Weighted Covariance Determinant Estimator (MWCD), was applied and compared with the multilayer neural network Back propagation to find the estimate of the median location based on the minimum distance (Mahalanobis Distance) and smallest specified for the variance matrix.

Joint Minimum Covariance Determinant (MCD) as one of the most nonparametric methods robust.

The study has been applied on environmental pollution statistics of drinking water for the year (2013) including all the Iraqi provinces except Kurdistan region, divided into (10) months.

To determine the contaminating indicators, Smoothing Spline slides were used to estimate the multivariate regression model parameters for the time – variable parameters.

The parameters smoothed by the cross validation (CV ) method were estimated the results of the comparison proved the effectiveness of the retrospective Multi – Layer neural network.