A hybrid training mechanism for applying neural networks to web-based applications
Ko-Kang Chu, Maiga Chang, Yen‐Teh Hsia · 2005
This paper proposes a hybrid training neural network and applying it to the accuracy counter (AC) developed previously. The neural network is used for detecting the cheating model for abnormal browsing behaviors performed by users in the conflicting environment. The most significant issue, training, should be taken into consideration while we are applying the neural network to Web-based applications such as the accuracy counter. Therefore, we design a hybrid Web based training mechanism for neural networks to deal with this kind of training problem. Finally, we also find out that the AC's block rate for detecting the abnormal browsing behaviors is increasing from 61% (rule-based) to 76% (neural networks with hybrid training mechanism) in the efficient and acceptable training period.