In today's digital world, most daily money transactions are done virtually through online systems. The rise in credit card transactions has increased the number of fraudulent transactions, leading to significant financial losses. Currently, the main problem faced during the analysis of transactions is the imbalance in the dataset. To address the issue of data imbalance and identify good models for accurately detecting fraudulent transactions, this paper presents a comparative study to determine the suitable machine learning algorithms for credit fraud detection. In this research study, Synthetic Minority Oversampling Technique (SMOTE) processing is done to balance the dataset, and various machine learning classifiers, Logistic Regression, Naive Bayes, K-Nearest Neighbor (KNN), Decision Trees, and Support Vector Machine (SVM) are compared and analyzed. During the experimental process, it was observed that after SMOTE was enhanced, SVM outperformed other models with an accuracy of 98.9%. When there are numerous features (variables) in the data, as is often the case in credit card transactions when several elements are taken into account, SVM can perform well. SVM differentiated outliers because of its margin-maximizing characteristics, which assisted in separating the fraudulent class from the non-fraudulent class.