International Journal of Applied Information Systems |
Foundation of Computer Science (FCS), NY, USA |
Volume 12 - Number 39 |
Year of Publication: 2022 |
Authors: T.O. Oyegoke |
10.5120/ijais2022451926 |
T.O. Oyegoke . A Hybrid Model for Classification of E-mail Fraud. International Journal of Applied Information Systems. 12, 39 ( April 2022), 13-24. DOI=10.5120/ijais2022451926
The study pre-processed e-mail data, formulated and validated a Particle Swarm Optimization (PSO)-based Back Propagation model for email fraud detection. This was done by the hybridization of two algorithms namely; Nature Inspired Algorithm and Artificial Neural Network. The dataset collected for the purpose of developing the model contained fraudulent mails (46.3%), Spam (32.6%) and Ham (21.1%) e-mails. 12,831 features were extracted after data preparation and cleaning, in which only 6,382 (49.7%) relevant features were selected using PSO. The model was simulated using 70% and 80% for training while 30% and 20% of datasets were used for testing respectively. The results of using the 30% and 20% testing dataset for the gradient-based BP algorithm showed that using the relevant features selected by PSO improved the accuracy by a value of 0.27% and 0.35% respectively while for the PSO-based BP algorithm, using the relevant features selected by PSO improved the accuracy by a value of 1.51% and 1.46% respectively. The results showed that using PSO-based BP had a better performance than gradient-based BP by a value of 1.48% and 2.72% for 30% training dataset and a value of 1.46% and 2.57% using the original features and the features selected using PSO respectively. The study concluded that the PSO-based BP algorithm was able to improve the performance of the Multi-Layer Perceptron compared to the Gradient-Based Back Propagation algorithm which has implications on improving advance fee fraud detection.