Supervised fraud detection of mobile money transactions on different distributions of imbalanced data : A comparative study of the classification methods logistic regression, random forest, and support vector machine
Sammanfattning: The purpose of this paper is to compare the classification methods logistic regression, random forest, and support vector machine´s performance of detecting mobile money transaction fraud. Their performance will be evaluated on different distributions of imbalanced data in a supervised framework. Model performance will be evaluated from a variety of metrics to capture the full model performance. The results show that random forest attained the highest overall performance, followed by logistic regression. Support vector machine attained the worst overall performance and produced no useful classification of fraudulent transactions. In conclusion, the study suggests that better results could be achieved with actions such as improvements of the classification algorithms as well as better feature selection, among others.
HÄR KAN DU HÄMTA UPPSATSEN I FULLTEXT. (följ länken till nästa sida)