Fixed potential security issues, including items proactively identified by vulnerability scanning software or PTC Quality Assurance testing. Please upgrade as soon as possible to take advantage of these important improvements.

For important recommendations about the use and update of third-party software, see technical support article CS320194.

Parameters have been added to the Profiles service so that non-mutually exclusive profiles can be generated. In other words, overlapping profiles can now be created, where the same record can be included in multiple profiles. Overlapping profiles can be generated via analytics services and in Analytics Builder.

This capability can reveal scenarios that may not be found with mutually exclusive profiles. The choice to generate overlapping profiles depends on the use case and type of data. In a case that includes redundant information across multiple features, mutually exclusive profiles will still produce more useful results.

TA-7547

New APIs are available to create and retrieve subpopulations of a dataset based on the results of a profiles job. This functionality provides a way to use the features of a specific profile to filter the dataset down to a subpopulation that includes only those records that match the profile.

The new create API can only run on a parquet dataset and overlapping profiles must be enabled. The new retrieve results API can be used to retrieve the create profile subpopulation results in an easily readable format. For more information about using the new APIs, see Profile Subpopulations.

TA-7791

An option has been added in the Training and Validation microservices to support the use of k-fold validation. This cross-validation technique can provide a more accurate estimate of how well a model created from the dataset will perform. This method is particularly useful for small datasets where reserving a holdout portion of data for validation could be undesirable.

When the new k-fold validation option is used in a training or validation job, a dataset is partitioned into a specific number of equally sized folds (smaller datasets). The methods by which a dataset is partitioned for k-fold validation depend on the type of data: non-time series, single-entity time series, multi-entity time series. For more information about dataset partitioning, see K-Fold Validation.

The results of k-fold validation jobs include a list of validation metrics from each k fold and statistics for each metric calculated across folds, including mean, standard deviation, min, and max.

In this release, Analytics Builder does not support k-fold validation.

TA-1130

New metadata parameters have been added to define how missing dataset values should be handled when an analytics job is run (signals, profiles, clusters, training, validation, or scoring). These new parameters include options for how the missing value should be treated, which values should be interpreted as missing, and (if applicable) what to replace them with. Methods for handling a missing value include:

For more information, see Handling Missing Dataset Values.

TA-7546

The values in dropdown lists for datasets, goals, and filters have been reordered to display in alphanumeric order. This change has been applied to screens for creating signal, profiles, models, or predictive scoring jobs. This enhancements is especially useful for finding values in large datasets.

TA-7272

New information about when a dataset was last updated has been added in multiple places. On the Analytics Builder Dataset page, this information is in a new column called Last Updated Date. In the ListCreatedDatasets service of the DataThing, the new column is labeled lastUpdatedDateAndTime.

In addition, in Analytics Builder, on-screen messages have been added to the Models, Signals, and Profiles pages to alert that the underlying dataset has been updated since the model, signal, or profiles job was created.

TA-7608

TA-4253/TA-4252/TA-4251

New validation has been added to ThingWorx Analytics services to check for dataset reference completeness. Some Analytics services require a dataset reference that includes both data and metadata information. If information is missing, an error message is displayed. This validation has been added to both Analytics Server and ThingWorx services.

TA-7990

Filtering functionality has been updated so that dataset filters are stored with the corresponding dataset in Analytics Server. New services have been added to the Data microservice to facilitate adding, retrieving, or deleting filters. With this enhancement, dataset filters are now available from Analytics Server APIs and Analytics Builder has been updated to access filters through these new services.

For more information about using the new filter APIs, see Handling Dataset Filters via ThingWorx Analytics DataThing.

TA-7561

TW-89203

Analytics Server and Platform Analytics have actually been using the TLS protocol since SSL was deprecated. However, install parameters and the installer UIs have continued to use SSL terminology. In this release, wherever possible, references to SSL have been updated to TLS. In the graphical and text mode installers, field names have been updated. In the silent mode installer, parameters names have been updated.

TA-7611

TA-7948

When a model is created using a large dataset (1 GB or more), the display of the model results can be slow. This issue has been resolved so that performance is improved and a work around is no longer needed to avoid a timeout failure.

TA-6766

In ThingWorx, if the Alert Processing Subsystem is not configured to point to the Analytics Server, anomaly alerts cannot be created. Previously, when an anomaly alert was created on a Thing, this connection was not validated during the alert creation. The anomaly alert still went through the calibration stage and then failed on training with an unclear message.

With this release, validation has been added, for Anomaly Detection in both standalone and cluster modes, so that proper configuration of the Alert Processing Subsystem is confirmed. Anomaly alerts will not be created if the configuration is invalid. Clear error messages have been added to the process.

TA-6973

TA-6976

In both Windows and Linux environments, when the regional format is set to German, anomaly alerts fail before they can reach the monitoring status. This issue has been resolved so that anomaly alerts operate successfully in German.

TA-7681

If the Redundancy Filter is enabled during predictive model creation, features are ranked according to the information gain they provide. The value in Max Fields limits the model to a specific number of fields during training. This Max Fields value is being ignored. This issue has been resolved so that the limits on the number of features included during training are observed.

TA-7788

In the results of a signals job in Analytics Builder, when you select a specific feature from the list on the left, the values in the grid on the right do not display as expected. This issue has been resolved so that all feature values display properly.

TA-7790

When modifying an existing anomaly alert on a Thing in ThingWorx Composer, an inaccurate error message displays saying that the selected Analytics Gateway does not exist. This issue has been resolved.

TA-8012

When a new signals job is defined with a Lookahead Size set to 0, the Lookahead Size value is changed to 1. This issue has been resolved so that a Lookahead Size value of 0 is not changed.

TA-8279

When a predictive scoring job is submitted via a REST API request, the information displayed on the Model list page in Analytics Builder is incorrect. The Number of Records Scored column should reflect the number of data rows submitted in the REST request. Instead, it displays a value of 0.

This issue will be resolved in a future release.

TA-4715

When a filter is applied to the Signals list page and results in no matching jobs, the list is not cleared properly in a Chrome browser and previous results are displayed. This issue will be resolved in a future release.

TA-4972

8012Analytics Manager: Empty analysis replay instance created

When an analysis replay is created and it does not include data for the specified time frame, an error is generated. However, an empty analysis replay instance is still created containing no records. The empty instance must be deleted before another replay can be created for the same time frame. This issue will be resolved in a future release.

TA-7619

This release includes new APIs to filter a dataset based on the features of a specific profile. The new APIs can be used to generate dataset subpopulations that includes only the records that match the features of a specific profile. See TA-7791 in the Enhancements section. However, a profile that includes features related to TEXT n-gram fields cannot be included when generating a subpopulation. This issue will be resolved in a future release.

TA-8106

When using a Java Edge SDK analysis agent in ThingWorx Composer, if you update a Thing property value and click Save to see the updated alert property values, the updated values may fail to display as expected. This issue will be resolved in a future release. As a work around, disable the analysis event and then re-enable it.

TA-8242

In ThingWorx Analytics Server 8.5, confidence models for Boolean and categorical goals could be created during training. When these confidence models are scored, the scoring output includes confidence interval information along with the prediction results. In a future release, the confidence interval information will no longer be returned as part of the scoring results for confidence models created in Analytics Server 8.5 with Boolean or categorical goals. In preparation for future upgrades, you should update any application code that is related to scoring these types of models. Ensure that your application code does not rely on the confidence interval results.