MACHINE LEARNING ALGORITHM FOR FRAUD DETENTION IN E-AUCTION SYSTEM

ABSTRACT

Virtual market place hosted on the internet is known as E-Auction. It is the process of buying and selling of items through online platforms. The bidder bids for the item, and the highest bidder is the winner of the item. Shill bidding occurs when fake bidders are introduced on the seller’s side to increase the final price. Shill bidding is a crime committed during the e-auction and it is pretty difficult to detect because it exhibits normal bidding behavior. The bidder gets a lot of loss because he pays extra money and the seller benefit from the process. The methodology adopted for the system analysis and design of the proposed system is object-oriented system analysis and design. This study proposed a neutral network-based model in which the model is been splitted into two parts, which are training and validation, where the training is 70% and validation 30%. The training set was used to train the model while the test set was utilized to evaluate the model performance. The proposed model used logistic regression to build the classifier to prevent shill bidding fraud in an online auction transaction. On every bid, it predicts whether the fraud is committed or not. If the bidding behavior is normal, continue the bidding; otherwise cancel the bid and block the user. The prediction accuracy of the proposed machine learning approach is 98.21%. The average error between the predictions and actuals in the data set is 0.1 which is fair enough since the variation is 1 out of 100. We compared the performance of the proposed model with matrices such as accuracy, precision and recall and also utilized visualization tools such as mat plot hub and seaborn to plot the result.