A comparison between classification statistical models and neural networks with application on Palestine data

Abstract

The paper has used labor force as dependent variable which contains two categories (Employment and Unemployment) and 8 independent variables. The results regarding the application of the correct classification technique to assess the accuracy of the three classification methods in predicting the labor force of have shown it was found that Artificial Neural Networks gave the best accuracy in prediction with (82.7%), 79.5% for Discriminant Analysis and (81.6%) for Logistic Regression. Furthermore, ROC curve technique has been applied to evaluate the accuracy of the three classification methods in predicting the labor force. It has been found that Artificial Neural Networks gave the best accuracy in prediction with (85.5%), (72.8%) for Discriminant Analysis and (81.7%) for Logistic Regression. In addition, Artificial Neural Network gave the best results in prediction with 82.7% accuracy, and less error rate with 0.173. Meanwhile, the Discriminant analysis model has shown 79.5% accuracy, and 0.205 error rate. Logistic Regression has shown 81.5% accuracy, 69.8% sensitivity and 0.183 error rate. These results demonstrate that Artificial Neural Network could be the most powerful analytical technique for the variables with two categories.