daily pattern AD

Thinkings:

1. Anomalous patterns may represent anomalous events. what does an anomalous pattern mean in energy consumption?
   1. measurement/manual operation errors, instrument failure, XX human activities. 比如窃电行为，偷入住行为。
2. What is the necessity of extracting anomalous knowledge?

　　　　1. they affect the models related to forecasting and analysis. 

　　　　2. The anomaly may be related to XX events. -----> early warning and prevention of XX disasters, and to reduce economic and social losses. 

　　3. What algorithms are used in energy domains? What algorithms are used in other fields? Can they be used in energy fields? if not, why?

1. other domains:
   1. SAX, convert TS into symbolic TS; then top-k pattern anomalies are analyzed and verified from the candidate pattern anomalies.
   2. formalize normal patterns, define a temporal similarity score between sequences. and then detect abnormal patterns which do not match or deviate from normal patterns. 
   3. learn a precise generative model of normal patterns and detect anomalies as cases that are not sufficiently explained by this model. 

　　4. Why anomaly detection is a challenging topic: 

　　　　1. mainly because of the insufficient knowledge and inaccurate representative of the so-called anomaly for a given system.

　　　　2. limited availability of labels to model AD as a discriminative classification task. 

　　　　3. due to massive data, which pose3 new challenges both computationally and statistically, and thus require novel approaches in discovering useful patterns. 

　　　　4. false alarm

　　5. How to construct a reliable anomaly detection method?　　

　　　　1. in Sequential Anomaly Detection using Inverse Reinforcement Learning: Rather than blindly taking the estimated reward as a normality score only, we also consider model confidence of the predicted values. 

• Anomaly detection definition: Anomaly detection, or outlier detection refers to (automatic) identification of unforeseen or abnormal phenomena embedded in a large amount of normal data.    Anomalies are referred to as abnormalities, deviants or outliers in the data mining and statistics literature