Higher Classification Accuracy of Income Class Using Decision Tree Algorithm over Naive Bayes Algorithm

Abstract

Developing two machine learning classifiers with higher accuracy for classifying income classes for people earning less and a higher salary scale between 50,000. Decision Tree Algorithm (DTA) and Naive Bayes Algorithm (NBA) are the two classifier mechanisms employed. On a dataset of 32516 records, the methods were implemented and tested. Implemented each algorithm through programs and performed ten rounds on both methods to determine distinct scales of income class for who earns lesser and higher salary scale between 50,000. The G-power test is around 80% accurate. The findings of the programming experiment showed that the Decision Tree Algorithm had a mean accuracy of 84.3790 and the Naive Bayes Algorithm had a mean accuracy of 79.3170 for classifying income categories. The variation in accuracy between the two classifiers is statistically significant (p=0.53), which is insignificant when employing the unpaired samples t-Test. The primary purpose of this work is to apply a novel technique to modern Machine Learning Classifiers to forecast income class classification. When the Decision Tree Algorithm is compared to the Naive Bayes Algorithm, the results show that the DTA outperforms the NBA.