anomalydetectionnode properties

Last updated: Jan 18, 2024

Anomaly node icon The Anomaly node identifies unusual cases, or outliers, that don't conform to patterns of "normal" data. With this node, it's possible to identify outliers even if they don't fit any previously known patterns and even if you're not exactly sure what you're looking for.

Example

node = stream.create("anomalydetection", "My node")
node.setPropertyValue("anomaly_method", "PerRecords")
node.setPropertyValue("percent_records", 95)
node.setPropertyValue("mode", "Expert")
node.setPropertyValue("peer_group_num_auto", True)
node.setPropertyValue("min_num_peer_groups", 3)
node.setPropertyValue("max_num_peer_groups", 10)

Table 1. anomalydetectionnode properties
`anomalydetectionnode` Properties	Values	Property description
`inputs`	[field1 ... fieldN]	Anomaly Detection models screen records based on the specified input fields. They don't use a target field. Weight and frequency fields are also not used. See Common modeling node properties for more information.
`mode`	`Expert` `Simple`
`anomaly_method`	`IndexLevel` `PerRecords` `NumRecords`	Specifies the method used to determine the cutoff value for flagging records as anomalous.
`index_level`	number	Specifies the minimum cutoff value for flagging anomalies.
`percent_records`	number	Sets the threshold for flagging records based on the percentage of records in the training data.
`num_records`	number	Sets the threshold for flagging records based on the number of records in the training data.
`num_fields`	integer	The number of fields to report for each anomalous record.
`impute_missing_values`	flag
`adjustment_coeff`	number	Value used to balance the relative weight given to continuous and categorical fields in calculating the distance.
`peer_group_num_auto`	flag	Automatically calculates the number of peer groups.
`min_num_peer_groups`	integer	Specifies the minimum number of peer groups used when `peer_group_num_auto` is set to `True`.
`max_num_per_groups`	integer	Specifies the maximum number of peer groups.
`num_peer_groups`	integer	Specifies the number of peer groups used when `peer_group_num_auto` is set to `False`.
`noise_level`	number	Determines how outliers are treated during clustering. Specify a value between 0 and 0.5.
`noise_ratio`	number	Specifies the portion of memory allocated for the component that should be used for noise buffering. Specify a value between 0 and 0.5.