Campaign | partitions | partition[n] | Interact | learning
These configuration properties enable you to tune the built-in learning module.
confidenceLevel
- Description
A percentage indicating how confident you want the learning utility to be before switching from exploration to exploitation. A value of 0 effectively shuts off exploration.
This property is applicable if the Interact > offerserving > optimizationType property for Interact runtime is set to BuiltInLearning only.
- Default value
95
- Valid Values
An integer between 0 and 95 divisible by 5 or 99.
validateonDeployment
- Description
If set to No, Interact does not validate the learning module when you deploy. If set to yes, Interact validates the learning module when you deploy.
- Default value
No
- Valid Values
Yes | No
maxAttributeNames
- Description
The maximum number of learning attributes the Interact learning utility monitors.
This property is applicable if the Interact > offerserving > optimizationType property for Interact runtime is set to BuiltInLearning only.
- Default value
10
- Valid Values
Any integer.
maxAttributeValues
- Description
The maximum number of values the Interact learning module tracks for each learning attribute.
This property is applicable if the Interact > offerserving > optimizationType property for Interact runtime is set to BuiltInLearning only.
- Default value
5
otherAttributeValue
- Description
The default name for the attribute value used to represent all attribute values beyond the maxAttributeValues.
This property is applicable if the Interact > offerserving > optimizationType property for Interact runtime is set to BuiltInLearning only.
- Default value
Other
- Valid Values
A string or number.
- Example
If maxAttributeValues is set to 3 and otherAttributeValue is set to other, the learning module tracks the first three values. All of the other values are assigned to the other category. For example, if you are tracking the visitor attribute hair color, and the first five visitors have the hair colors black, brown, blond, red, and gray, the learning utility tracks the hair colors black, brown, and blond. The colors red and gray are grouped under the
otherAttributeValue
, other.
percentRandomSelection
- Description
The percent of the time the learning module presents a random offer. For example, setting percentRandomSelection to 5 means that 5% of the time (5 out of every 100 recommendations), the learning module presents a random offer, independent of the score. Enabling
percentRandomSelection
overrides theofferTieBreakMethod
configuration property. WhenpercentRandomSelection
is enabled, this property is set regardless if learning is on or off or if built-in or external learning is used.- Default value
5
- Valid Values
Any integer from 0 (which disables the
percentRandomSelection
feature) up to 100.
recencyWeightingFactor
- Description
The decimal representation of a percentage of the set of data defined by the recencyWeightingPeriod. For example, the default value of .15 means that 15% of the data used by the learning utility comes from the recencyWeightingPeriod.
This property is applicable if the Interact > offerserving > optimizationType property for Interact runtime is set to BuiltInLearning only.
- Default value
0.15
- Valid Values
A decimal value less than 1.
recencyWeightingPeriod
- Description
The size in hours of data granted the recencyWeightingFactor percentage of weight by the learning module. For example, the default value of 120 means that the recencyWeightingFactor of the data used by the learning module comes from the last 120 hours.
This property is applicable only if optimizationType is set to builtInLearning.
- Default value
120
minPresentCountThreshold
- Description
The minimum number of times an offer must be presented before its data is used in calculations and the learning module enters the exploration mode.
- Default value
0
- Valid Values
An integer greater than or equal to zero.
enablePruning
- Description
If set to Yes, the Interact learning module algorithmically determines when a learning attribute (standard or dynamic) is not predictive. If a learning attribute is not predictive, the learning module will not consider that attribute when determining the weight for an offer. This continues until the learning module aggregates learning data.
If set to No, the learning module always uses all learning attributes. By not pruning non-predictive attributes, the learning module may not be as accurate as it could be.
- Default value
Yes
- Valid Values
Yes | No