Predicting Accuracy of Income a Year Using Rough Set Theory

• Main Author: Zuraihah, Ngadengon
• Format: Thesis
• Language: English; English
• Published: 2009
• Subjects: QA273-280 Probabilities. Mathematical statistics
• Online Access: http://etd.uum.edu.my/2066/1/Zuraihah_Ngadengon.pdf; http://etd.uum.edu.my/2066/2/1.Zuraihah_Ngadengon.pdf; http://etd.uum.edu.my/2066/

The main objective of the experiments is to predict the accuracy of Adult dataset whether the income exceeds $50K per year or below $50K. Specifically, the objectives are to determine the best discretization method, split factor, reduction method, classifier and to build the classification model. In the experiments, the prediction of accuracy of the Adult dataset is developed by using rough set theory and Rosetta software while Knowledge Data Discovery (KDD) is used as the methodology. The Adult dataset that had been used in the experiments is comprises of 48,842 instances but only 24,999 instances is used along the experiments. Then, the data was randomly split into training data and testing data by using nine splits factor, which are 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8 and 0.9. The result obtained from the experiments showed that the best discretization method is Naive Algorithm, the best split factor is 0.6, the best reduction method is Johnson's Algorithm and the best classifier is Standard Voting. The highest percentage of accuracy achieved by the classification model developed using the rough set theory is 87.12%. The experiments showed that rough set theory is a useful approach to analyze the Adult dataset because the accuracy achieved in the experiments exceeds the previous methods that have been used before.