DLOT-Net: A Deep Learning Tool For Outlier Identification

Abstract

Outlier identification is one of the trending research projects, which is used to detect the normal (important) and abnormal (abusive, unimportant, attack) content presented in the data. So, the automatic outlier detection plays the major role in various applications. However, the conventional methods are failed to provide the maximum accuracy, efficiency due to ineffective classification. Therefore, this work focused on implementation of deep learning-based outlier tool network (DLOT-Net). Initially, Outlier Detection Datasets (ODDS) is considered for simulations, which is preprocessed to remove the missed symbols. Then, the deep learning convolutional neural network (DLCNN) model trained with the preprocessed dataset. During the training process, DLCNN model creates the memory of outliers. Then, for every random test sample, the DLCNN model identifies the normal and abnormal attributes presented in the data using probability comparisons. The simulations conducted on ODDS dataset shows that, the proposed DLOT-Net resulted in superior objective performance as compared to several other outlier detection methods.