DATA ANALYSIS APPLICATION AND NAÏVE BAYES SUPERVISED CLASSIFICATION IN ONLINE PAYMENT

Abstract

The fast development of online payment transactions has led to an increase in fraud in this type of transaction, causing great losses for many individuals and collectives in the financial industry. Credit transaction fraud in online payment is one of the most common and disturbing illegal activities. The detection, prevention of fraudulent transactions through analysis and data mining combined using machine learning algorithms is one of the current prominent methods. Data mining techniques are used to study patterns, characteristics, attributes and behaviors of normal transactions, abnormal transactions (fraudulent transactions) based on standardized and irregular data. Class machine learning algorithm to predict, detect normal transactions, fraudulent transactions automatically whenever a new transaction arises. This paper looks at some supervised machine learning algorithms: Using Bayes network, Tree Augmented Naïve Bayes (TAN) and Naïve Bayes in the binary classification problem based on data are more than 4 million online credit transaction records equivalent to about 80,000 card codes to detect fraudulent transactions. After pre-processing the data using the Principal Component Analysis (PCA) method, all classification algorithms achieve 95% more accuracy than the pre-pretreated data set.