SUMMARY
In this study a model which can improve the accuracy and reliability of credit card fraud detection was proposed. This is with a few to mitigating contentious issues regarding online transaction of credit card, such as amount of transactions that have resulted in payment default and the number of credit card fraud cases that have been recorded, all of which have put the economy in jeopardy. To address this challenge,sample dataset was sourced from online repository database of Kaggle. The feature extraction on the data was performed using Principal Component Analysis (PCA). The credit card fraud detection model was designed using Neuro-fuzzy logic technique, clustering was done using Hierarchical Density Based Spatial Clustering of Application with Noise (HDBSCAN) .The simulation of the proposed model was done in Python programming environment.The performance evaluation of the model was carried out by comparing the proposed model with Neuro-Fuzzy (NF) technique using performance metrics such as precision, recall, F1-score and accuracy. The simulation result showed that the proposed model (NF + HDBSCAN) had precision of 98.75%, recall of 98.70%, F1-Score of 97.65% and accuracy 99.75% . NF had Precision of 94.60%, recall of 94.50%, F1-Score of 95.50% and accuracy 95.70% using training dataset. Likewise, when test dataset were used, the proposed (NF + HDBSCAN) had precision of 93.50%, recall of 95.50%, F1-Score of 94.50% and accuracy 95.50%. NF had Precision of 92.50%, recall of 93.00%, F1-Score of 94.00% and accuracy 93.50%. The simulation results of the proposed model was viable, reliable and showed possibility of being designed as module which could be integrated into the existing credit card design for lowering fraud rate and assisting fraud investigators.