In recent years fraudulent transactions have become a major concern for the finance industry due to the amount lost every year. Manually analyzing the transactions and finding out the fraud transactions is a tedious task. Hence machine learnings algorithms can learn the patterns from the data and detect fraudulent transactions. But the significant challenge here is imbalanced data where non-fraudulent transactions are very dominant in the dataset. To overcome this problem, we will be using oversampling and under-sampling techniques. To detect the fraudulent transaction in the dataset we will be using ML algorithms like Logistic Regression, Naïve Bayes and Random Forest with sampling techniques
a)Logistic Regression:-
b)Naive Bayes:-
c)Random Forest:-
In this paper, we have performed the experiments using Logistic Regression, Na¨ıve Bayes, and Random Forest to classify the fraud transactions with sampling techniques to overcome the imbalanced data problem. From the results obtained in Section with metrics like precision, re-call and f-score Random forest with or without any sampling technique has performed better than other algorithms mentioned in this paper. Another important observation from the experiments is that sampling techniques have increased the performance of all the algorithms. From the metrics, we can observe that after the random forest classifier Logistic regression has better performance.