Predictive Intelligence properties

  • Release version: Xanadu
  • Updated August 1, 2024
  • 2 minutes to read
    • Summarize
    Summarized using AI
    This content was generated using new OpenAI-powered functionality. Results are provided on an as is basis and are not guaranteed to be accurate or complete.

    Summary of Predictive Intelligence properties

    The Predictive Intelligence properties in ServiceNow control key aspects of machine-learning frameworks, solution training processes, and other application features. These properties can be reviewed or edited viaPredictive Intelligence > Configurationusing themaintrole, with some requiring additional roles for access. Adjusting these settings allows customers to tailor the behavior and limits of machine learning solutions such as classification, regression, similarity, and clustering.

    Show full answer Show less

    Key Properties and Their Practical Use

    • Training Lock Override: The glide.platformml.overridetraininglock property enables editing of machine learning solution definitions during training by overriding their readonly state.
    • Regression Training Limits: Properties glide.platformml.api.minregressionrecords and glide.platformml.api.maxregressionrecords set minimum and maximum record counts for regression training. Note that support for creating new regression solutions was deprecated after the Washington DC release, though existing ones can still be trained and edited.
    • Training Timeout: The glide.platformml.trainingtimeout defines how long (in milliseconds) a training session can wait without updates before timing out, defaulting to 6 hours.
    • Model Size and Record Limits:
      • glide.platformml.api.modelsize limits the maximum number of records used to train a model.
      • glide.platformml.api.csvmaxline and glide.platformml.api.csvminline set the allowed record range for classification training.
      • glide.platformml.api.csvsplitdays caps the maximum number of days’ worth of records retrievable per request.
    • Word Corpus and Similarity Window Limits:
      • glide.platformml.api.maxwordcorpusrecords limits records per table for Word Corpus creation, though this is mostly obsolete for clustering and similarity since these now use pre-trained Workflow solutions.
      • glide.platformml.api.maxsimilaritywindowrecords limits the number of records retrieved for similarity lookups.
    • Clustering Record Limit: The glide.platformml.api.maxclusteringrecords determines the maximum records included in clustering solutions.
    • Shared Service Scheduler: The URL for shared service scheduling is set by glide.sharedservicescheduler.url.
    • Machine Learning Artifacts Management:
      • glide.cache.size.mlobjectcache controls the maximum cache size for ML artifacts in MB.
      • glide.platformml.artifact.cachecompressionscheme specifies the compression scheme used for artifact caching.

    Why This Matters

    By configuring these properties, ServiceNow customers can optimize machine learning performance, manage data volume constraints, and control training behaviors to better fit their operational needs. This ensures efficient use of resources and compliance with platform updates such as deprecation of certain solution types.

    What to Expect

    Customers can expect more granular control over their Predictive Intelligence models, including training durations, data size limits, and artifact caching. They should be aware of deprecations (e.g., regression solutions) and adjust their workloads accordingly. Proper configuration supports more reliable and scalable predictive analytics within the ServiceNow platform.

    The Predictive Intelligence properties control certain aspects of its machine-learning solution frameworks and solution training process, as well as other areas of the application.

    Using the maint role, navigate to Predictive Intelligence > Configuration to review or edit these properties.
    Note:
    Some properties require other roles to review or edit.
    Table 1. Predictive Intelligence properties
    Property Property Name Description
    Override ml_solution_definition record's readonly state during training glide.platform_ml.override_training_lock
    • True: Overrides the readonly state during training, enabling you to edit ml_solution_definition.
    • False: ml_solution_definition remains in the readonly state.
    Minimum number of records for Regression glide.platform_ml.api.min_regression_records Sets the minimum number of records required for Regression solution training.
    • Type: Integer
    • Default value: 10000
    Note:
    Support for new regression solutions is deprecated with the Washington DC release. You can edit and train any existing solutions, but you can't create new ones.
    Maximum number of records for Regression glide.platform_ml.api.max_regression_records Sets the maximum number of records that can be used in Regression solution training.
    • Type: Integer
    • Default value: 300000
    Note:
    Support for new regression solutions is deprecated with the Washington DC release. You can edit and train any existing solutions, but you can't create new ones.
    The time (in ms) an in-training solution will wait without updates before timing out glide.platform_ml.training_timeout Sets the time-out time in milliseconds for an in-training solution.
    • Type: Integer
    • Default value: 21600000 (ms)
    Maximum model size glide.platform_ml.api.model_size Sets the maximum number of records you can use to train a model.
    • Type: Integer
    • Default value: 524288000
    Maximum number of records used in training glide.platform_ml.api.csv_max_line Sets the maximum number of records that can be used in Classification solution training.
    • Type: Integer
    • Default value: 300000
    Minimum number of records used in training glide.platform_ml.api.csv_min_line Sets the minimum number of records required for Classification solution training.
    • Type: Integer
    • Default value: 10000
    Maximum number of days worth of records on request can retrieve glide.platform_ml.api.csv_split_days Sets the maximum number of days a single request can retrieve from your records.
    • Type: Integer
    • Default value: 30
    Maximum number of records per table for word corpus glide.platform_ml.api.max_wordcorpus_records Sets the maximum number of records per table for Word Corpus creation for Similarity and Clustering solutions.
    • Type: Integer
    • Default value: 300000
    Note:
    With the Washington DC release, clustering and similarity solutions use Workflow solutions. These are pre-trained so a word corpus isn't needed. The Word Corpus field is removed from Workflow solution forms.
    Maximum number of records for similarity window (to lookup results) glide.platform_ml.api.max_similarity_window_records Sets the maximum number of records that the Similarity Window can retrieve to look up results.
    • Type: Integer
    • Default value: 100000
    Maximum number of records for Clustering glide.platform_ml.api.max_clustering_records Sets the maximum number of records you can include in a cluster.
    • Type: Integer
    • Default value: 100000
    Table 2. Shared Service Scheduler
    Property Property Name
    Shared service scheduler url glide.shared_service_scheduler.url
    Table 3. Machine Learning Artifacts
    Property Property Name
    Maximum number of artifacts cached (in MB) glide.cache.size.ml_object_cache
    Artifact cache compression scheme glide.platform_ml.artifact.cache_compression_scheme