Research on imbalanced data set preprocessing based on deep learning

Feature mining for encrypted malicious traffic detection with deep learning and other machine learning algorithms

“…Zhang et al [6] proposed a transfer learning method based on Efficientnet to detect encrypted malicious traffic. The author first pre-trained an Efficientnet model, Efficientnet-B0, through the imagenet dataset, and then transfer it to a small amount of encrypted traffic dataset for training.…”

“…The authors defined three modes of extracting flows: Application flows, conversation flows, and End-point flows. Zhang et al [6] proposed a novel encoding method for protocol-specific traffic features for TLS/SSL protocols. The encoding method converts the extracted features into an image-like data format.…”

“…Such an imagelike feature set is fed into CNN models. Although the features they created in [11] [6] are specific features for encrypted traffic, only TLS/SSL/IPSec protocols are considered. Once the traffic contains other encrypted protocols, such features are no longer applicable.…”

Strategies of Predictive Schemes and Clinical Diagnosis for Prognosis Using MIMIC-III: A Systematic Review

“…Apart from this, it is also stated at the beginning of this section about the need to maintain consistency, appropriateness, and effective structurization of clinical data using machine learning. This is where there is a demand created for applying preprocessing operations over any existing machine learning approaches on clinical datasets, as stated in the work of Aljuffri et al [46], Zelaya [47], and Fangyu et al [48]. In this perspective, a recurrent neural network is found to sort out this problem to some extent and is quite applicable in learning-based data structurization where current computation is dependent on prior outcomes.…”