Autoencoder

Autoencoder is a type of artificial neural network used for unsupervised learning and anomaly detection. It is widely used for time series forecasting outlier detection because of its ability to model complex, non-linear patterns in the data.

Autoencoder Parameters (OutlierTechId = 5)

Param Name	Description	Default Values	Possible Values
EPOCHS	The number of times the entire dataset is passed through the model during training.	30	<=500
IQR-MULTIPLIER	Multiplier for the Interquartile Range (IQR) used for outlier detection.	1.5
LR	The learning rate that controls the size of weight updates during training.	0.001
USE-RECONSTRUCTION	Whether to use the reconstruction error to detect outliers.	True
PATIENCE	The number of epochs with no improvement in validation loss before training stops.	5