Analysis of Accuracy for Approving Bank Loan using Numerical Data of Customer by Comparing Decision Tree over Logistic Regression Algorithm

Abstract

Aim: To predict the loan of the person using a Novel Decision Tree (DT) over the Logistic Regression algorithm (LR). Materials and Methods: The existing model uses a Logistic Regression algorithm and in the proposed Novel Decision Tree. The 20 sample values are used to find out the mean. Std. deviation, std. error means. The sigmoid function is calculated using Levene’s Test for Equality of Variances both assumed and non-assumed. Decision trees use different algorithms to decide to split a node into two or more sub-nodes. The purity of the node depends upon the target variable. The prediction of the loan can be done only after knowing all the factors. The sample size was measured as 40 for both groups using G power (80%). Results: The graph explains the comparison of the mean accuracy value with algorithms Novel Decision Tree and Logistic Regression where the mean accuracy is about 88% and 70%. The statistical significant value obtained is 0.78 (p>0.05), it shows insignificance between the groups based on an independent sample T-Test. Conclusion: The mean accuracy rate of the Novel Decision tree algorithm has been improved to 88% compared to Logistic Regressionwhich is having around 70% mean accuracy.