Analyzing and Optimizing Results

When you have obtained your first results in Rulex, you will probably want to analyze and optimize them.

There are may tasks that allow you to do just this on the many structures produced by Rulex tasks.

For further information

Structures can also be converted into datasets and vice-versa to perform in-depth analysis operations.

For more information on this see the section Converting Structures

Available tasks

Task Name

Structure

Description

Corresponding Page

Rule Manager

Rules

Inspects, manipulates and optimizes sets of rules.

Input Tasks

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

Analyzing Rules in the Rule Manager

Adding Rules Manually in the Rule Manager

Rule Viewer

Rules

Graphically displays rules, and retrieves information such as their importance, the relationship between classes of output attributes with respect to input attributes.

Input Tasks

This task can only be used with rules generated by the Classification LLM task.

Analyzing Classification LLM Rules in the Rule Viewer

Feature Ranking

Rules

Graphically displays the importance of attributes within a class values within specific attributes.

Input tasks

The task can be used only with rules that originate from one of the following tasks:

Ranking Rule Features and Values

Merge Rules

Rules

Merges rules from multiple computations.

Merging Rules

Optimize Ruleset

Rules

Improves the generation of predictive rules through a series of constraints.

Optimizing Rulesets

Association Manager

  • Association rules

  • Replacement rules

Analyzes association and replacement rules, using filtering or sorting operations, computing statistics or creating plots.

The task is very similar to the Data Manager, but is designed specifically for replacement and association rules.

Optimizing Association and Replacement Rules

Find/Replace

Rules

Replaces values that can be modified with new values to improve the outcome of the analysis.

Finding and Replacing Values in Datasets

Itemsets/Sequences Manager

  • Frequent Itemsets

  • Frequent Sequences

Analyzes, filters and sorts frequent itemsets and sequences in a similar way to the Data manager task. 

Task input

You must have produced clusters, itemsets or sequences in your process via one of the following tasks:

Analyzing Frequent Itemsets and Sequences

Confusion Matrix

Model

Computes and visualizes the performance of any classification method.

Input tasks

The task can be used with any classification task, but must be proceeded by an Apply Model task that tests the classification results on data.

Analyzing Model Performance in the Confusion Matrix

Apply Model

Model

Applies a model generated in Rulex on data.

Input tasks

Models can be made up of classification and regression rules and clusters, and the task can consequently be used after any classification, regression or clustering task.

Applying Models to Data