استخدام طرق التصنيف لتحديد أهم عوامل الخطر على مرضى السكري في فلسطين-قطاع غزة

Abstract EN

Diabetes considers one of the most prevalent diseases; there are many research dealt with this topic.

but this research plays an important role as it was able to build a statistical model which can predict the cases of diabetes, that we can identify the most important risk factors which the more appear on a person, the far more likely to be infected with the disease.

Three statistical models (Multiple Logistics Analysis, discriminant analysis, neural network analysis) were tested to obtain the best model can classify diabetes as a dependent variable and the most important risk factors that help the presence of this disease.

The conclusion was that the best model we got is by using (neural network analysis) as the accuracy of classification and prediction (95.7%).

The most important risk factors were: (1) high psychological stress (2), quantity of fruits and vegetables consumed per week, (3) amount of meat consumed per week, (4) scientific qualification.

In this research, the researchers used some statistical methods for model evaluation and for evaluating the quality of the classification achieved by this model: (leave-one-out cross validation, classification table and ROC curve).

Data in this research has been analyzed using software R, SPSS.