large e-commerce company with its operations in several countries. As the online giant grows, so has the number of fraudster merchants are. They deliver counterfeits or, in some cases, nothing at all. Such schemes leave customers duped, and place both legitimate merchants and the company itself in a constant battle to rid the marketplace of scammers. Determining this is also important in budgeting for fraud investigation. It's a well-known problem both to the company and to merchants, which they say hasn't effectively addressed the issue. They are serious about it and want to protect themselves from these fraudulent merchants using technology.
Created an analytical and modelling framework to predict the Merchant Fraudulency(yes/no) based on the quantitative and qualitative features provided in the dataset while answering other questions too cited "fraudester" -Yes =1 , No = 0.
Practically the same was a highly imbalance data where chances of detection can take place due to lack of evidence,hence to achieve this developed profiling of important features of legitimate customer behavior and indicated anomalies.
A Meta-classifier model was used in this research. This model consist of 3 base classifier constructed using Decision Tree, Random Forest and Naive Bayesian Algorithms.
Results from this research show that when a meta-classifier was deployed in series , Performance improvement has been changed to 24% to 36% which can be resulting in savings of million $.