Anomaly detection in a network based on a key performance indicator prediction model
摘要
說明書
As shown, input layer 126 receives values of the KPI and associated parameters at a plurality of times as inputs to neural network 124. If a value of the KPI at one or more of the plurality of times is missing, the neural network may be used to predict the missing values of the KPI, and use the prediction as inputs to neural network 124. Neural network 124 may use the intermediate layers (e.g., hidden layers) to determine the prediction of the particular value of the KPI based on the parameters associated with the particular value of the KPI. For example, the intermediate layers may include one or more feedforward layers and/or one or more recurrent layers to determine the prediction for the particular value of the KPI. The one or more feedforward layers and/or recurrent layers may include a plurality of coupled nodes that are linked according to being trained as described herein. In this way, links between nodes of intermediate layers 128 may correspond to predictions, classifications, and/or the like that are associated with the parameters that would lead to determining the prediction that is within a threshold range (or within a threshold level of accuracy) of a non-anomalous KPI value at time T. In some implementations, the output layer 130 may include a predicted KPI value at time T. in some implementations, the output layer 130 may include a range of non-anomalous values of the KPI at time T (e.g., between a lower anomaly threshold and an upper anomaly threshold).
