Fraud detection in shipping industry based on location using machine learning comparison techniques

Thumbnail Image
Ganesan Subramaniam, Mr.
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
This research discusses fraud detection specifically within the shipping industry using data analytics techniques. Shipping industry volume in general is skyrocketing due to the rapid pace of globalization and the significant e-commerce growth fueled by the recent pandemic. As it grows strongly it also attracts fraudsters to find their ways to commit fraudulent shipping arrangements such transporting illegal or banned items using fake declaration documents. Inspecting shipments and verifying related documents for a huge volume of shipments task is becoming such an impossible task and it’s no longer sustainable as the volume continues to surge in locations all around the world. Shipping companies and customs are mostly relying on routine random inspection thus finding this fraud is often by chance. This issue will also increase operational cost for the shipping industry if more resources are to be deployed to handle this task manually. Manual task will also cause longer shipping processing time that would be impact the efficiency of the supply chain which in turn lead to increase of transportation cost which finally leads overall cost of goods to consumers. As it’s no longer sustainable and effective for both shipment companies and customs to pursue traditional fraud detection strategies such there needs to be an automated solution before the problem impacts the industry and increase crime rates which will cause a bigger impact to society as a whole. Other related papers on this area have proven that intelligent data-driven fraud detection is proven to be far more effective than routine inspections. However, the challenge in data-driven detection is its effectiveness is often detection accuracy and the speed of detection. Accuracy of detection is often impaired by various fraud mechanisms used by fraudsters to commit shipment. Speed of detection derived from the speed of model execution is also important for earlier detection of fraudulent cases. As such in this study, reviews were conducted and subsequently the most optimized model with existing algorithms combination were identified to detect fraud effectively within the shipping industry. Criteria for the effectiveness for the shipping industry would be detection performance in terms the time it takes to identify the fraud and also the accuracy of the detection. There were also identification of factors that influence fraud activity, review existing fraud detection models, develop the detection model and implement it using a well-known tool in the market namely Rapidminer. A number of popular existing algorithms were used to execute the model developed in Rapid tool such as Naïve Bayes , Neural Net , Deep Learning, Decision Tree, Logistic Regression, SVM and k-NN. Results from this study indicates the usage of k-NN is proven to be most effective algorithm in terms of performance and accuracy required by the shipping industry fraud detection.
Fraud detection in shipping industry