Impact of Classification Algorithms on Census Dataset

Abstract

Data mining is a method by which valuable information can be obtained from large databases. A supervised method of classification assigns data samples to target groups. In this system, it uses various classification algorithms namely decision trees, SVM, random forest and neural network. This system will classify and analyses the best suited algorithm which gives maximum accuracy among the other algorithms. The accuracy in these algorithms has been calculated by sensitivity and specificity. Evaluation of these models has been calculated by the error rate with respect to the classes. It uses census dataset and finds whether the income above 50k or below 50k. Matrix of error consists of true positive, neutral, true negative and false negative values. Based on true positive and false negative values, specificity is determined. Based on true negative and false positive values, sensitivity is determined. The algorithm analysis which finds the better algorithm with respect to the accuracy, error rate and efficiency.