Hybrid Algorithm for Naïve Bayes-Based Credit Card Fraud Detection

Many fraud transactions exist in the online world that affects various financial institutions but Credit Card Fraud transaction is the most occurring problem in the world. Credit Card fraud is the situation in which fraudsters misuse credit cards for illegal purposes. Hence, detection of fraudulent transactions is essen-tial. Several researchers have worked on detecting fraud transactions and also provide solutions whose surveys are given in this paper. This study makes a major contribution to research on the detection of Credit Card fraud transactions through Machine Learning Algorithms suchas Decision Tree and Naive Bayes. The data have been selected from Kag-gle and categorize into training (80%) and testing (20%) data. The whole experiment was performed on the Jupyter Notebook tool for which the Anaconda Navigator has been installed. The Heatmap is used for visualization and colorfully represents the data. The main aim of this work is to balance the dataset with Near-Miss Under-sampling Method. The information gain method is applied for feature selection. The best algorithm founded in this paper is Decision Tree with 97% accuracy as compared to Naïve Bayes with 90%. The results are achieved based on Accuracy, Recall, Precision, and F1-score. We have also shown the ROC Curve and Precision-Recall Curve of the algorithm in this paper.

Download Full-text

Application of Credit Card Fraud Detection Based on CS-SVM

International Journal of Machine Learning and Computing ◽

10.18178/ijmlc.2021.11.1.1011 ◽

2021 ◽

Vol 11 (1) ◽

pp. 34-39

Author(s):

Chenglong Li ◽

◽

Ning Ding ◽

Haoyun Dong ◽

Yiming Zhai ◽

...

Keyword(s):

Support Vector Machine ◽

Random Forest ◽

Credit Card ◽

Naive Bayes ◽

Search Algorithm ◽

Fraud Detection ◽

Cuckoo Search ◽

Cuckoo Search Algorithm ◽

Support Vector ◽

Credit Card Fraud

With the development of e-commerce, credit card fraud is also increasing. At the same time, the way of credit card fraud is also constantly innovating. Support Vector Machine, Logical Regression, Random Forest, Naive Bayes and other algorithms are often used in credit card fraud identification. However, the current fraud detection technology is not accurate, and may cause significant economic losses to cardholders and banks. This paper will introduce an innovative method to optimize the support vector machine by cuckoo search algorithm to improve its ability of identifying credit card fraud. Cuckoo search algorithm improves classification performance by optimizing the parameters of support vector machine kernel function (C, g). The results demonstrate that CS-SVM is superior to SVM in Accuracy, Precision, Recall, F1-score, AUC, and superior to Logistic. Regression, Random Forest, Decision Tree, Naive Bayes, whose accuracy is 98%.

Download Full-text

Ensemble Classification Method for Credit Card Fraud Detection

International Journal of Recent Technology and Engineering - 2 ◽

10.35940/ijrte.c4213.098319 ◽

2019 ◽

Vol 8 (3) ◽

pp. 423-427

Keyword(s):

Execution Time ◽

Credit Card ◽

Naive Bayes ◽

Research Work ◽

Hybrid Approach ◽

Fraud Detection ◽

Naïve Bayes ◽

Ensemble Classification ◽

Bayes Classifier ◽

Proposed Model

Credit card frauds are on the rise and are getting smarter with the passage of time. Usually, fraudulent transactions are conducted by stealing the credit card. When the loss of the card is not noticed by the cardholder, a huge loss can be faced by the credit card company. In the existing work, it has been found that the researchers have utilized Voting based method to identify credit card frauds. The problem with voting based method is that they are more complex and more time consuming. In this research work, a hybrid approach based on KNN and Naive Bayes for the detection of credit card frauds. KNN will be used as the base classifier and it will return predicted result. The predicted result will be provided as input to the Naive Bayes classifier which will generate the final result. The proposed model will be compared with existing techniques and the results are analyzed in terms of recall, precision, accuracy and execution time.

Download Full-text