1.
Anomaly Detection algorithm based on time series predictionThe statistical method based on data Data Model is the most extensive Anomaly Detection technology. Its basic principle is to statistically model the training Dataset (usually normal samples). If a data sample does not conform to the random Data Model obtained from training, it will be identified as a Exception sample. The fact that the data sample does not conform to the random Data Model means that the sample is unlikely to be generated by the Data Model. It is generally determined through statistical testing, based on the following assumption: normal data samples occur in the high probability area of statistically random Data Model, while the data sample of Exception occurs in the low probability area of Data Model. In other words, we can get the probability that the unknown data sample is generated by the known Data Model based on the test statistic. If the probability is less than a certain predetermined standard, the sample is considered to be Exception.
2.
Intelligent alert algorithm based on machine learningThrough the Anomaly Detection algorithm based on time series prediction, we can allow the alert system to automatically and relatively accurately complete fault detection and alert work, thereby greatly reducing the arduous manual setting of alert thresholds for various performance indicators and manual management tasks, and achieving the purpose of automatic and intelligent fault detection.
