An Efficient Communication Intrusion Detection Scheme in AMI Combining Feature Dimensionality Reduction and Improved LSTM

Table 7

The comparison of results between the proposed method and traditional machine learning methods.

Method	Accuracy	Precision	Recall	FAR
K-nearest neighbor	0.7544	0.7942	0.6977	0.1870
Naive Bayesian	0.8358	0.7399	0.8731	0.1869
Decision tree	0.8531	0.7811	0.8765	0.1624
Random forest	0.8583	0.8434	0.8400	0.1268
Logistic regression	0.8471	0.7212	0.9190	0.1917
Support vector machines	0.6486	0.9964	0.5599	0.0079
Multilayer perceptron	0.8095	0.7147	0.8350	0.2063
SAE + Attention-BiLSTM	0.9941	0.9914	0.9952	0.0069