Comparing the Performance of Decision Tree and Artificial Neural Network in the Classification of National Exam Candidates

Comparison of capabilities of three different algorithms of decision tree and neural network model in order to choose the appropriate classification model to evaluate the performance of national exam candidates.

It is based on a quantitative approach and a survey method, and the variables of gender, academic records and balance scores of each course were used as effective variables in the classification.

Considering all the variables without gender based on the neural network model, it was found that the specialized courses of mathematics, physics and chemistry, followed by the general courses of Persian and religion respectively, are the most important in the classification. The distance between the diploma and the national exam had the least effect. It was also found that if only variables related to test courses are considered, the order of importance of chemistry and physics courses will be shifted.

Based on the overall accuracy index, the neural network model algorithm with an accuracy greater than 0.95 has a higher performance than the decision tree algorithms. On the other hand, the inclusion of educational background variables has had a favorable effect on the accuracy of the neural network algorithm.