Table 6 CoLog in comparison to other log anomaly detection methods on the Nssal dataset. Bold numbers indicate the outstanding results.

Supervised Methods

Logistic Regression¹⁴³

85.133

74.728

76.476

97.604

Support Vector Machines¹⁰⁹

80.206

74.935

76.474

97.655

Decision Tree¹¹⁰

94.791

87.700

89.470

98.063

Attentional BiLSTM⁶³

96.750

98.805

97.754

99.813

Convolutional Neural Network⁴⁶

96.703

98.243

97.460

99.789

pylogsentiment³⁸

97.170

96.050

96.602

99.020

Unsupervised Methods

Isolation Forest¹⁴⁴

65.504

57.352

56.101

80.967

Principal Component Analysis¹²

52.642

53.827

49.505

80.614

LSTM³⁶

96.148

97.669

96.896

99.742

Transformer⁸⁰

96.304

99.354

97.778

99.813

\(\hbox {CoLog}^{1}\)

\(\mathbf {99.955}\)

\(\mathbf {99.915}\)

\(\mathbf {99.935}\)

\(\mathbf {99.967}\)