Accounting Analysis by Incorporating Apriori Association Rule Algorithm

Abstract

Electronic payment systems are increasingly being used globally for online business transactions. Accounting analysis provides domain expertise to assist feature selection, preprocessing, and model interpretation, hence improving anomaly identification in credit card data using deep learning. Deep learning models can effectively identify anomalies, detect fraudulent activities, and continuously improve fraud detection capabilities in credit card transactions by integrating accounting principles and financial knowledge. This guarantees strong fraud prevention measures for financial institutions. In this manuscript, Accounting Analysis by Incorporating Apriori Association Rule Algorithm (AA-IAARA-PMNN-PCBESA) is proposed. Initially, the input datas collected from credit card dataset are given as input. The input datas are fed to pre-processing using Confidence Partitioning Sampling Filtering (CPSF) for identifying the missing values from the input data. Afterward the pre-processed datas were given to feature selection using Black Winged Kite Algorithm (BWKA) for selecting the transaction features. Then selected features were given to Port-Metriplectic Neural Network (PMNN) optimized with Polar Coordinate Bald Eagle Search Algorithm (PCBESA) for accurate detection of anomaly in the credit card data and classify the detected anomaly as fraud and non-fraud. The proposed AA-IAARA-PMNN-PCBESA approach is implemented in Python. The performance of the proposed AA-IAARA-PMNN-PCBESA  technique attains19.5%, 24% and 23% higher accuracy, 24.6%, 23.15% and 24.8% higher Precision and 21.4%, 27.36% and 21.08% higher recall compared with existing methods such as a Performance Evaluation of Machine Learning Methods for Credit Card Fraud Detection Using SMOTE and AdaBoost (PE-CCFD-SVM) ,machine learning based credit card fraud detection using the GA algorithm for feature selection (CC-FD-RF) ensemble learning approach for anomaly detection in credit card data with imbalanced and overlapped classes (AD-CCD-XGB) models respectively .