Analyzing Rules in the Rule Manager

The Rule Manager allows you to inspect, manipulate and optimize a set of rules.

It can be used with any of the following LLM tasks that generate rulesets:

Prerequisites

Additional tabs

The following additional tabs are provided:

  • Documentation tab where you can document your task,

  • Parametric options tab where you can configure process variables instead of fixed values. In this task the parametric options are available for alert options only.

  • History tab, which tracks all  the operations that have been performed in the Rules Manager task, and behaves in the same way as the Data Manager History tab, with icons for the main operations:

    • Minus, for the deletion of rules or conditions

    • Plus, for the creation of new rules or conditions

    • Question mark, for query operations


Procedure

  1. Drag and drop the Rule Manager task onto the stage.

  2. Connect a task, which contains the ruleset you want to analyze, to the new task.

  3. Double click the Rule manager task. The left-hand pane displays how many rules have been generated and the percentage of these total rules are currently displayed in the ruleset. This percentage may change if modifications are made, such as applying filters or displaying rules only with selected attributes.

  4. Filter results as described in the table below.

  5. Save and compute the task.

Rule Manager options

Parameter Name

Description

Select rules for output

Filter the ruleset that will be used in the output by:

  • #Conditions - specify the number of conditions each rule can contain.

  • Covering - specify the covering percentage that filtered rules must respect.

  • Error - specify the error percentage that filtered rules must respect.

Select rules containing

Select the attributes which must be included in the rules. Only those rules whose conditions contain at least one of the selected attributes or output attributes will be displayed.

Search attribute

Search for attributes by quickly retrieving them from the list.

Order attributes by

Sort attributes by attribute order, name or type.

Sort conditions by

Sort conditions according to their attributes, covering or error values.

Filter conditions

Filter conditions so that any attributes that have not been selected in the Select rules containing lists are removed.

Results

The rule analysis is divided into three separate spreadsheets:

Rules spreadsheet

The Rule spreadsheet, with the generated ruleset, contains the following columns:

  • #Cond: the number of conditions in the rule.

  • Output: the output if that rule is matched.

  • Cond n: the n-th condition.

In this spreadsheet you can modify rules and conditions:

  • To delete a rule, select its row in the table and either click the minus icon above the spreadsheet or right-click and select Delete selected rules.

  • To add new rules either click the plus icon above the spreadsheets or by right-click within the spreadsheet and select Create rule. Once you have specified the output value for your rule, a blank rule will be added to the table to which you will then need to add/append conditions.  

  • To delete a condition select it in the spreadsheet, right-click and select Delete selected condition or simply press Delete. The conditions to be removed can belong to different rules. 

  • To edit a condition either select it in the spreadsheet, right-click and select Edit condition or simply double-click it. The changes you can make depend on whether the condition is an ordered or nominal condition.

  • To add or append a new condition to a rule, double-click an empty cell in the row, or right-click and select Append condition, then select an attribute from the dialog box:



Covering spreadsheet

The Covering spreadsheet, with additional information related to the covering of each rule in the following columns:

  • #Patt.: the number of patterns in the training set with the same output class of the rule.

  • Covering: the percentage of patterns (with the output class of the rule) matched by this rule in relation to the total number of patterns of that class.

  • w\o Cond n: the covering gain that would be obtained by removing the n-th condition.

Error spreadsheet

The Error spreadsheet, with additional information about the error scored by each rule in the following columns:

  • #Patt.: the number of patterns in the training set with the output class different from the output class of the rule.

  • Error: the percentage of patterns (with the output class different from that of the rule) matched by this rule in relation to the total number of patterns where the output class is different from the class of the rule.

  • w\o Cond n: the error increase that would be obtained by removing the n-th condition.