Hybrid cnn attention model for accurate ad click fraud detection

Abstract

The internet scam of clicking is a bane to online advertisers whose campaigns data is being fabricated. Attention- based convolutional neural network will form the basis of the paper in order to enhance the deep learning models. It enables this model to nullify the majority of most noteworthy time and behavior features of sequential data of the click streams. They are geared towards identifications accuracy improvements. In cases where the sequential dependency learning and contextual representation are upgraded, the attention-enhanced CNNs would be effective in fraud classification as compared to the normal CNNs. The reason is that this is done to achieve real-time prediction in a web application developed with the Flask based on the proposed model. Dynamic online advertising ecosystems are successful, useful and effective in nature.