Hybrid ML Classifier for Loan Prediction System

Abstract

Banks make the majority of their income from loans. A lot of individuals apply for loans, and it is difficult to choose the real candidate who will repay the loan. A lot of misunderstandings may occur when selecting the real applicant when the process is done manually. As a result, a loan prediction system based on machine learning is developed, in which the system will automatically identify the qualified candidates. This is beneficial to both the bank personnel and the applicant. The loan approval process will be greatly shortened. The loan data is predicted by using the hybrid model of Naive Bayes (NB) and Decision Tree (DT) algorithms. First, the dataset is given to the three classification algorithms– Support Vector Machine (SVM), NB and DT Algorithms and the prediction is done with these three algorithms. The accuracy of each of these three is used to assess performance. The creation of the hybrid model increases accuracy. The dataset is given to NB for training and the prediction of NB is given to DT Algorithm for training. Test data are sent to the model for prediction after training. The model is evaluated, and the performance is measured in terms of different metrics form sklearn metrics. This prediction of loan range is useful for bank staff to give the loan amount accordingly. The NB algorithm checks for equality and independence of all the features in the dataset. In DT algorithm, the tree is constructed based on the information gain value. The attribute with high information gain value is placed as the root node and also the other nodes are constructed based on information gain value. The proposed hybrid model predicts - yes or no, and based on the prediction, whether the loan is to be sanctioned or denied for the applicant is specified.