SIMANJUNTAK, JEFRY ROBERTO HASUDUNGAN (2023) COMPARISON OF K-NEAREST NEIGHBORS (KNN) AND NAÏVE BAYES ALGORITHMS FOR FRAUD DETECTION OF CREDIT CARD TRANSACTIONS. Other thesis, UNIVERSITAS KHATOLIK SOEGIJAPRANATA.

Preview	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - COVER_a.pdf Download (631kB) | Preview
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - BAB I_a.pdf Restricted to Registered users only Download (239kB)
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - BAB II_a.pdf Restricted to Registered users only Download (537kB)
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - BAB III_a.pdf Restricted to Registered users only Download (460kB)
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - BAB IV_a.pdf Restricted to Registered users only Download (838kB)
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - BAB V_a.pdf Restricted to Registered users only Download (185kB)
Preview	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - DAPUS_a.pdf Download (219kB) | Preview
	Text 19.K1.0057-Jefry Roberto Hasudungan Simanjuntak - LAMP_a.pdf Restricted to Registered users only Download (443kB)

Abstract

This study aims to evaluate the performance of the KNN and Naïve Bayes algorithms in Detection of credit card transactions. Two datasets, Data 1 and Data 2, were utilized, each employing different outlier handling techniques. The experimental results highlight the performance of the KNN and Naïve Bayes algorithms for each dataset. For Data 1, missing values were removed, while for Data 2, outlier values were eliminated. The following results were obtained for Data 1 (missing value removal) are KNN algorithm in Data1 KNN Oversampling (20:80) achieved an accuracy of 98.98. Naïve Bayes algorithm in Data1 Naivebayes (30:70) achieved an accuracy of 95.16%.And for Data 2 (outlier removal) are KNN algorithm in Data2 KNN Oversampling (20:80) achieved an accuracy of 99.79%. Naïve Bayes algorithm in Data2Biasa Multinomial Naïve Bayes (30:70) achieved an accuracy of 99.31%. The findings indicate that the KNN algorithm demonstrates robustness in handling datasets with numerous outliers. It effectively addresses data imbalance through oversampling and undersampling techniques. On the other hand, Naïve Bayes exhibits limitations in handling datasets with a high number of outliers and performs better when the dataset contains fewer outliers. This research provides insights into the effectiveness of the KNN and Naïve Bayes algorithms for outlier handling in data classification, contributing to the selection of appropriate algorithms for handling specific types of datasets.

Actions (login required)

View Item