Computer Science

Permanent URI for this collection

http://dspace.alquds.edu:4000/handle/20.500.12213/1091

Browse

Browsing Computer Science by Author "AMJAD A. M. HARB"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    دراسة مقارنة الأداء والكفاءة بين تقنيات التنقيب عن البيانات والأساليب الإحصائية في تصميم نماذج التنبؤ
    (AL-Quds University, 2012-05-10) امجد عبد المنعم محمود حرب; AMJAD A. M. HARB; رشيد الجيوسي; Nidal Kafri; Yousef Abuzir
    The aim of this research is to perform a comparison study between statistical and data mining modeling techniques. These techniques are statistical Logistic Regression, data mining Decision Tree and data mining Neural Network. The performance of these prediction techniques were measured and compared in terms of measuring the overall prediction accuracy percentage agreement for each technique and the models were trained using eight different training datasets samples drawn using two different sampling techniques. The effect of the dependent variable values distribution in the training dataset on the overall prediction percent and on the prediction accuracy of individual “0” and “1” values of the dependent variable values was also experimented. For a given data set, the results shows that the performance of the three techniques were comparable in general with small outperformance for the Neural Network. An affecting factor that makes the percent prediction accuracy varied is the dependent variable values distribution in the training dataset, distribution of “0” and “1”. The results showed that, for all the three techniques, the overall prediction accuracy percentage agreement was high when the dependent variable values distribution ratio in the training data was greater than 1:1 but at the same time they, the techniques, fails to predict the individual dependent variable values successfully or in acceptable prediction percent. If the individual dependent variable values needed to be predicted comparably, then the dependent variable values distribution ratio in the training data should be exactly 1:1.