Table 6 Comparison of anomaly detection methods in literature
From: Time series big data: a survey on data stream frameworks, analysis and algorithms
Type of Method | Method | PM | Year | Metrics | |
|---|---|---|---|---|---|
Statistical | Based on recursive least squares, and sparsity maximization | [75] | 2016 | F-Score, ROC, Residual error | |
| Â | Â | Based on wavelet filters and pseudo-spline filters | [59] | 2002 | TP |
| Â | Â | Based on correlation techniques | [97] | 2016 | Absolute error |
| Â | Â | Based on Dirichlet process | [77] | 2019 | Accuracy, FPR, TPR |
| Â | Â | Based on seasonal decomposition and robust statistical metrics | [58] | 2017 | F-Score, TPR, Precision |
ML | Based on PCA | Based on rPCA | [61] | 2017 | FPR, FNR |
|  |  | Based on PCA and the Karhunen Loève Expansion | [62] | 2013 | AUC, ROC |
| Â | Â | Based on multi-scale analysis, PCA, and wavelet transforms | [60] | 2015 | ROC |
| Â | Based on KNN and TCM | [63] | 2009 | FPR, TPR | |
| Â | Naive Bayes | [36] | 2018 | Accuracy | |
| Â | Based on SVM | SVM | [22] | 2015 | Accuracy |
| Â | Â | Based on RBM and SVM | [70] | 2019 | Accuracy, FPR, F-Score, ROC, Precision |
| Â | Based on SOM | SOM | [71] | 2005 | FPR, TPR |
| Â | Â | SOM with k-medoids | [37] | 2018 | FPR |
| Â | Based on tensor factorization | [76] | 2017 | FPR, TPR | |
DL | Based on FNNs | [64] | 2019 | Accuracy, Error rate, FPR, F1-Score, Precision, TPR | |
| Â | Based on RNNs | Based on GRU | [66] | 2021 | Accuracy, F1-Score, Precision, TPR |
| Â | Â | Based on RNNs | [67] | 2017 | Accuracy, AUC, FPR, Loss, ROC, TPR |
| Â | Â | Based on LSTM | [65] | 2018 | AUC, ROC |
| Â | Based on CNNs | Based on CNNs | [68] | 2018 | TPR |
| Â | Â | Based on CNNs | [74] | 2018 | MCC |
| Â | Â | Based on CNNs and FNNs | [57] | 2018 | Accuracy, FPR, TPR |
| Â | Â | Based on CNNs | [69]* | 2020 | Accuracy |
| Â | Based on Autoencoders | Based on Autoencoders and convolution | [72] | 2018 | Accuracy, FPR, ROC |
| Â | Â | Based on Stacked Autoencoders | [73] | 2019 | Accuracy |