A Comparative Analysis of Traditional and Deep Learning-Based Anomaly Detection Methods for Streaming Data

Munir, Mohsin; Chattha, Muhammad Ali; Dengel, Andreas; Ahmed, Sheraz

doi:10.1109/icmla.2019.00105

Cited by 34 publications

(17 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Providing a comparison statement between the aforementioned methods, deep learning achieves better results, in terms of accuracy and loss, in contrast with other traditional machine learning methods, in cases where the input data size increases over time, as the considered case where the OBD sensors from the vehicles continuously send data. Furthermore, the evolution of parallel computing methods with the integration of powerful GPUs that leads to high-end infrastructures, provides higher capacity and computational power, key features in order to achieve better training results, in terms of time [ 42 , 44 ], when deep learning techniques are applied. Another major issue that needs to be taken into account during the selection of machine or deep learning methods is the existence and the importance in the first case (machine learning methods) of a hardcore feature preprocessing in order to drop specific, useless features with a view to reducing the input dataset and eventually reduce the complexity of the procedure.…”

Section: Dataset and Algorithms/methodsmentioning

confidence: 99%

Driving Behaviour Analysis Using Machine and Deep Learning Methods for Continuous Streams of Vehicular Data

Πεππές

Alexakis

Adamopoulou

et al. 2021

Sensors

View full text Add to dashboard Cite

In the last few decades, vehicles are equipped with a plethora of sensors which can provide useful measurements and diagnostics for both the vehicle’s condition as well as the driver’s behaviour. Furthermore, the rapid increase for transportation needs of people and goods together with the evolution of Information and Communication Technologies (ICT) push the transportation domain towards a new more intelligent and efficient era. The reduction of CO2 emissions and the minimization of the environmental footprint is, undeniably, of utmost importance for the protection of the environment. In this light, it is widely acceptable that the driving behaviour is directly associated with the vehicle’s fuel consumption and gas emissions. Thus, given the fact that, nowadays, vehicles are equipped with sensors that can collect a variety of data, such as speed, acceleration, fuel consumption, direction, etc. is more feasible than ever to put forward solutions which aim not only to monitor but also improve the drivers’ behaviour from an environmental point of view. The approach presented in this paper describes a holistic integrated platform which combines well-known machine and deep learning algorithms together with open-source-based tools in order to gather, store, process, analyze and correlate different data flows originating from vehicles. Particularly, data streamed from different vehicles are processed and analyzed with the utilization of clustering techniques in order to classify the driver’s behaviour as eco-friendly or not, followed by a comparative analysis of supervised machine and deep learning algorithms in the given labelled dataset.

show abstract

Section: Dataset and Algorithms/methodsmentioning

confidence: 99%

Driving Behaviour Analysis Using Machine and Deep Learning Methods for Continuous Streams of Vehicular Data

Πεππές

Alexakis

Adamopoulou

et al. 2021

Sensors

View full text Add to dashboard Cite

show abstract

“…RF would be used for nonlinear regression tasks [13]. As a research example, in Munir et al, several methods and modeling approaches have been mentioned to detect and classify anomalies, which is considered statistical modeling for conducting anomaly classification process [14]. For instance, K nearest neighbor (KNN) method is one of the most commonly used distance-based methods for anomaly detection.…”

Section: Related Studiesmentioning

confidence: 99%

An LSTM-Autoencoder Architecture for Anomaly Detection Applied on Compressors Audio Data

Mobtahej

Zhang

Hamidi

et al. 2022

Computational and Mathematical Methods

View full text Add to dashboard Cite

The compressors used in today’s natural gas production industry have an essential role in maintaining the production line operational. Each of the compressors’ components has routine maintenance tasks to avoid sudden failures. Hence, the significant advantages and benefits of performing preventative maintenance tasks in time are decreasing equipment downtime, saving additional costs, and improving the safety and reliability of the whole system. In this paper, anomaly classification and detection methods based on a neural network hybrid model named Long Short-Term Memory (LSTM)-Autoencoder (AE) is proposed to detect anomalies in sequence pattern of audio data, collected by multiple sound sensors deployed at different components of each compressor system for predictive maintenance. In research methodology, this paper has conducted experiments that employed different RNN architectures such as GRU, LSTM, Stacked LSTM, and Stacked GRU with various functions to create a baseline for model evaluation. Each architecture used audio signals dataset received from the compressor system for experiments to consider each neural network model’s performance. According to performance results, an optimal model for anomaly detection with the best performance scores has been proposed in this research. Experiments combined one-dimensional raw audio signal features using SC and Mel spectrogram features were fed to deep learning models to evaluate performance. Hence, such hybrid methods can effectively detect normal and anomaly audio signals collected from a compressor system, increasing the compressor system’s reliability and the sustainability of the gas production line. The combination of multiple-resource features in the proposed hybrid model showed a 100% score in all four-evaluation metrics such as accuracy, precision, recall, and F1 in LSTM-based autoencoder in both test and train results.

show abstract

“…Isolation forest is a notable, decision tree-based exception [22]. These approaches, however, suffer from issues of scaling in both data volume and complexity [3], [4], [23].…”

Section: Related Workmentioning

confidence: 99%

Modified Autoencoder Training and Scoring for Robust Unsupervised Anomaly Detection in Deep Learning

Merrill

Eskandarian

2020

IEEE Access

View full text Add to dashboard Cite

The autoencoder (AE) is a fundamental deep learning approach to anomaly detection. AEs are trained on the assumption that abnormal inputs will produce higher reconstruction errors than normal ones. In practice, however, this assumption is unreliable in the unsupervised case, where the training data may contain anomalous examples. Given sufficient capacity and training time, an AE can generalize to such an extent that it reliably reconstructs anomalies. Consequently, the ability to distinguish anomalies via reconstruction errors is diminished. We respond to this limitation by introducing three new methods to more reliably train AEs for unsupervised anomaly detection: cumulative error scoring (CES), percentile loss (PL), and early stopping via knee detection. We demonstrate significant improvements over conventional AE training on image, remote-sensing, and cybersecurity datasets.

show abstract

A Comparative Analysis of Traditional and Deep Learning-Based Anomaly Detection Methods for Streaming Data

Cited by 34 publications

References 25 publications

Driving Behaviour Analysis Using Machine and Deep Learning Methods for Continuous Streams of Vehicular Data

Driving Behaviour Analysis Using Machine and Deep Learning Methods for Continuous Streams of Vehicular Data

An LSTM-Autoencoder Architecture for Anomaly Detection Applied on Compressors Audio Data

Modified Autoencoder Training and Scoring for Robust Unsupervised Anomaly Detection in Deep Learning

Contact Info

Product

Resources

About