What's New in Rulex 4.3

Principal Component Analysis task

NEW

The Principal Component Analysis task identifies the most important components in a dataset and consequently reduces the number of attributes a dataset contains.

The task basically compresses a large amount of data into a more manageable dataset that captures the essence of the original dataset.

Convert Structure to Dataset and Convert Dataset to Structure tasks

NEW

Two new tasks have been added to Rulex to make it possible to convert structures to datasets and vice versa. These tasks can convert all data structures, from models, to rules, including PCA eigenvectors and discretization cutoffs.

The previous more specific tasks (Convert Ruleset to Dataset, Convert Model to Dataset, etc.) have been left for compatibility, but from now on it will be sufficient to convert structures with these new tasks, and simply select the structure type in the task options.

Discretize task

CHANGE

A new option has been added to the Discretize task, which enables you to use the cut-off points created in an upstream Discretize task, instead of creating new cut-offs each time the task is used in the workflow.

Import from Task

CHANGE

Import from Task has been enhanced to include the new structures available in Rulex 4.3 (discretization cutoffs and PCA eigenvectors), and to define the existing structures in a clearer way.

Select Flows task

CHANGE

New structures have been included in the Select Flows task:

• cut-offs defined in Discretize tasks

• eigenvectors defined in Principal Component Analysis tasks

• models created in Auto Regressive tasks.

Export tasks

CHANGE

All export tasks in Rulex now include the following new structures, which can be exported to the target specified by the task (database, Microsoft Excel, JSON file, text file):

• cut-offs defined in Discretize tasks

• eigenvectors defined in Principal Component Analysis tasks

• models created in Auto Regressive tasks.

Export to Database task enhancements

CHANGE

New options have been added to the Export to Database task to improve the management of missing values and complex conditions:

• The management of missing values has been enhanced to provide increased flexibility on how missing values are interpreted in conditions. This option is necessary in conditional export operations, such as Update, Insert/Update and Delete.

• Complex conditions, such as time spans in dates or ranges of continuous values, can be manually defined as rules, which are then fed to the task as input. The new option allows you to define that you want to use these input rules as conditions in the task. 

All export to database operations (except Insert/Update) are now performed in bulk mode, considerably speeding up all operations.

LLM

CHANGE

A new option has been added to the LLM tasks (classification, regression and one-task) to limit the number of nominal values a condition may contain. This option simplifies resulting conditions, and avoids the risk of over-fitting when conditions are excessively precise for the test set data.

The calculation of relative and absolute errors has also been improved, along with the way the classes generated in LLM tasks are then applied in the Apply Model task.

Workflow Versioning enhancement

CHANGE

Two new options have been added to the workflow versioning feature:

• database repository details can be modified for a single process by right-clicking the specific process in the Processes pane, and selecting Version control > Change repo.

• two processes can be merged, using the merge policy options to resolve conflicts.

API Server

CHANGE

A new API has been added, which enables the deletion of processes via API. This is useful to avoid filling up Rulex working databases with old processes, which we no longer need.

Two computation modes have been added to the API server that were previously only available via CLI. The addition of these computation modes aims to align the two interfaces:

• non-inmemory 

• store sources