A distributed installation is now available via a Mesosphere Data Center/Operating System (DC/OS) cluster. The distributed installation decouples the microservices that accept job requests and queries from the common workers that execute the submitted jobs. In a distributed deployment, the microservices are lightweight and consume fewer system resources. The workers, which require more system resources, can handle any type of job request. If multiple common workers are deployed, multiple jobs of the same type can be processed simultaneously.

However, to facilitate this extra orchestration, a distributed deployment requires some additional infrastructure (such as ZooKeeper, PostreSQL, shared file system, and DC/OS components). Setup and configuration of the necessary components brings some complexity to the distributed installation process and requires that the user be comfortable working in a Linux command line environment. The distributed installation is only available by request from the ThingWorx Analytics Technical Support team.

A native Windows installer has been added to the lineup of options available for the standalone server. Standalone installation is now provided via a Docker installer tool (for either Windows or Linux environments) and via non-Docker installer tools for both Windows and Linux. Each installer bundles the microservices and other libraries into a series of files. When the installation process is launched, each tool unpacks and installs the necessary components.

This release includes the ability to return some new evaluation metrics that improve the usefulness of the model evaluation service. The new statistics include industry standard evaluation techniques for both Boolean and Continuous goal variables. All of the new metrics are exposed to the Analytics Server APIs, including the following:

TA-32

The new metrics added in this release to improve the usefulness of the model evaluation service (see above), are also exposed to the Analytics Builder user interface. The new statistics are included on the Model Results pages, as appropriate (based on the goal variable).

In addition, when a model that was created in an earlier ThingWorx Analytics release is selected for viewing, no values are available for these new evaluation metrics. Dashes (- -) are displayed instead.

TA-34

Optional upsampling and downsampling strategies have returned for training on Boolean goals in this release. These sampling strategies can be used to balance the true/false outcomes in a set of training data. Downsampling will decrease, by a user-defined factor, the number of records that match either the true or false value. Conversely, upsampling will increase the number of records that match the selected value by a user-defined factor.

The new upsampling and downsampling options are exposed to the Analytics Server API via the Training microservice. The strategies are applied globally to all of the learners included in a training ensemble technique.

TA-793

The new upsampling and downsampling strategies added in this release to help generate balanced training data (see above) are also exposed to the Analytics Builder user interface. The new parameters are available on the Advanced Model Configuration tab of the New Predictive Model dialog box.

The new parameters include a Strategy (upsample, downsample, or none), A Value to upsample or downsample on (true or false – for Boolean goals only), and a Factor by which to up or down sample the records that match the selected value.

TA-798

The following changes have been made that effect the installation process for both Docker and non-Docker installers:

In addition, the installation guides for both Docker and non-Docker standalone installers have been added to the ThingWorx Analytics Help Center. These guides are now available from both the Help Center and the PTC eSupport Portal Reference Documents.

—

With this release, the following changes have been implemented:

—

In a Windows operating system, when using the Docker Toolbox, it may appear that the installed ThingWorx Analytics Docker containers are not accessible via the Docker Quickstart Terminal, even though DockerMachinePTC is running. This scenario can occur when the containers were installed in Administrative mode, but the Quickstart Terminal is launched in default mode. One solution is to launch the Quickstart Terminal in Administrator mode.

However, if the Quickstart Terminal was already launched in default mode, it can be difficult to launch in Administrator mode. The server will need to be rebooted or the VirtualBox process restarted. An alternate solution is to use the Windows Power Shell, also in Administrator mode, instead of the Docker Quickstart Terminal.

Also, in some scenarios, the VirtualBox Manager might show that DockerMachinePTC is stopped when it is actually running. To avoid this issue, best practice is to use the Windows Power Shell, in Administrator mode, to view the machine status.

TA-203

Previously, the Op Type, INFORMATIONAL, was not included as an available option when creating a data filter. As a result, display-only fields that are typically assigned the INFORMATIONAL Op Type, could not easily be included in data filtering. For example, multiple steps were required to use an INFORMATIONAL Op Type field like record purpose to separate training data from scoring data. This issue has been resolved.

TA-708

CS273475

In the previous release, true positive rates (TPR) and false positive rates (FPR) were calculated incorrectly in validation statistics. This error has been corrected so that:

TA-944

When ThingWorx server is installed on the local server, and localhost is entered for connection purposes during the ThingWorx Analytics Docker installation, the connection validates successfully but the ThingWorx Analytics Server Things are not created in ThingWorx. The Edge Microserver in the Docker container cannot accept localhost as the connection to ThingWorx. The following possible work arounds are available to resolve this issue:

TA-536

When running a request for Signals, and specifying maxAtATime = 2, Mutual Information (MI), in relation to the goal, is not returned for all field combinations as expected. Instead, individual (one-at-a-time) MI scores are returned for all fields and then are filtered down to the top 25% most relevant fields. Two-at-a-time signals are calculated only for those fields. There is currently no way to modify this filtering behavior.

TA-729

When a dataset filter is deleted, all models, signals, profiles, and scoring jobs created using that filter should also be deleted. Currently, these items are not all being removed. New jobs cannot be initiated using the deleted filter, but if a user tries to retrain a model that was based on a deleted filter, the job will run against the entire dataset.

TA-502

When a time series model is built using the default lookback window size of 0, the lookback size is not retained when the time series model is retrained. This issue can manifest itself in two scenarios:

As a work around, rerun the training as a new model, using the same configuration as the original training job.

TA-472

While creating an analysis model for ThingPredictor, if you specify a network location in the Model File URL field, ThingPredictor cannot download the file and deploy the model.

AM-807

When deployer agent deploys an agent on a machine, the corresponding job status is shown to be in the INPROCESS state. However, the deployment is completed.

AM-819

Analysis agents must be restarted after a ThingWorx server restart. Without restart of the agent, any new simulation jobs will fail to execute.

—

The ContinuousModelEvaluator calculates Pearson Correlation and RMSE in different bins of the validation dataset. However, when the results are returned, their order is swapped and the values are mapped to the wrong keys. As a result, Pearson Correlation values can exceed 1 and RMSE results can be returned as negative values. Because of this incorrect mapping, the normalized RMSE values are actually a normalization of the Pearson Correlation, which is not a relevant result. The incorrect mapping will be resolved in a future release.

TA-1628