Table 6 Benefits and weaknesses of machine learning algorithms
Reference | Benefits | Weaknesses |
|---|---|---|
Eldali et al. [73] | They expose the REFIT dataset with detailed annotation for anomaly electricity consumption. | The dataset excludes uncommon anomalous features, which led to an increase in MSE. The application of loss estimation and theft detection is not sufficient. |
Cui and Wang [72] | The method improves the speed of detection effectively. | Changing the acquisition frequency constantly may cause harm to smart meters. The over-fitting challenge brings out a high false positive. |
da Silva et al. [74] | The classification method detects anomaly detection in early time as for real-time data. | The method has lower detecting accuracy. The slight electronic abnormal consumption owns the great impossibility to be detected. |
Liu et al. [75] | The detection method does not affect the operation of smart home devices. | It is difficult to make accurate detection when users are using non-smart devices by this method. |