A thesis Submitted to College of Administration and Economic-Karbala University in partial fulfillment of the Requirements for the Degree of master of Science in statistics
By
Murtada Farhan H. AL – Shuwaili
Under supervision
Asis. Prof. Dr. Shrooq A.S.AL-Sabbah
Abstract
Statistical forecasting remains one of the most important statistical analyzes since it is impossible to deny the need for accurate systems for reliable forecasting for future management and decision-making , and because future decision-making is accompanied by a degree of risk there for prediction in turn reduces the size of these risks wide range of models and styles.
From the previous studies and research, the idea of studying a real pathological phenomenon and analyzing its data using two different methods and comparing them to find the best method or model between them to study the phenomenon in the resigned and similar phenomena in terms of data pattern. In the theoretical side of this study The method of Artificial Neural Networks (ANN) and Cox Regression Model (Cox) and Kaplan-Meyer Method.
A random sample (186) with a major thalassemia was as survival data. The Kaplan-Meyer method was then used to determine the significance of independent variables of 14 variables and then used Artificial Neural Networks (ANN) as a new method for analyzing and comparing this type of data With Cox Regression Model for this type of data, and two cases for both methods. The first case included all variables of the study and the second case included the seven variables resulting from the Kaplan-Meyer method.
The artificial neural network selected in the second case, which includes only the structural variables (2-3-7), the training ratio 70%, and the test ratio of 30%, proved its advantage over its counterpart, which included in its analysis all the variables of the study or the moral ones, , As is the case with the Cox regression model. The results of the estimated model of the moral variables only are the best. This indicates the change that the Kaplan-Meyer method has made in terms of its effect on the efficiency of the method of statistical analysis and the extraction of the moral variables.
When comparing the results of the selected methods, it is found that the traditional methods are not always the best. The Artificial Neural Networks (ANN) proved its efficiency and efficiency at the expense of Cox Regression Model the two methods were compared with the correct prediction and the prediction rate, Use of this study in the statistical program (SPSS) and language (R)) and the program (Excel).
