8.4 Visualizing Association Sets

Association sets return a set of predicted items given some input data for a single instance. As we explained in section 8.3 , association sets computes the similarity score between the input data and the rule’s Antecedent from the original association model. Then, for rules with a similarity score greater than zero, the Consequent part of the rule is returned as a predicted item (as long as the items in the consequent are not part of the input data).

There are two cases in which you will not obtain any predicted items for your input data:

• If the input data is not found among the rules discovered in the original association model. Then the following warning message will be displayed:

  

• If the original association model contains only categorical and numeric fields and you set values for all of them. This is due to the fact that categorical and numeric fields only have one single value per instance. So if they are given as input fields, they cannot be returned as predicted items at the same time. For example, if you already set age=20 as input, BigML will not return the age as output since a person cannot have two different ages at the same time. In those cases the following warning message will be displayed:

  

BigML provides two different views for your predicted items, the table and the diagrams explained below.

8.4.1 Association Set Table

The table contains the rules from the original association model which antecedent matches the input data. (See Figure 8.14 .)

The consequent part of the rules contains the predicted items associated to their score. (See Figure 8.15 .) See section 8.3 for a full explanation of the consequent score.

The table also contains the measures for each of the matching rules: Coverage, Support, Confidence, Leverage or Lift. (See Figure 8.16 .) See section 2.1 for a detailed explanation of each measure.

BigML displays up to four rules in the same view. To view more rules, use the pagination options at the bottom of the table. (See Figure 8.17 .)

You can also filter the rules by typing any item or field name within the search box or using the measure slider. (See Figure 8.18 .)

Finally, export the table in CSV format by clicking on the option show in Figure 8.19

8.4.2 Association Set Diagrams

When you select a rule from the table, you will see two Venn diagrams displayed above the table. The association diagram on the left indicates the actual intersection between the antecedent and consequent itemsets of this rule, and the intersection if unrelated diagram on the right indicates the intersection if both itemsets were independent. These diagrams provide a visual overview of the importance of the selected association rule. See section 5.1 for a full explanation.

You can hide or show this view by clicking in the corresponding option. (See Figure 8.21 .)