Association Discovery with the BigML Dashboard

5.1 Associations Table View

After associations are created, you will get a table (Figure 5.2 ) that summarizes all the rules discovered.

\includegraphics[]{images/assoc-table-overview}
Figure 5.2 Associations table overview

At the top part, from left to right, you can see some basic descriptive information, such as the number of instances contained in this dataset (2,832), the number of item fields (32), number of associations discovered (100), the search strategy chosen (leverage), the percentage set for the minimum support (4.0610%), and the percentage set for the minimum confidence (23.6290%).

\includegraphics[]{images/table-top}
Figure 5.3 Descriptive information

Below this descriptive information, you can filter your associations by typing the items or values in the input box, or by moving the slider to filter rules by the measure used as the search strategy, leverage in this example. (See Figure 5.4 .)

\includegraphics[]{images/table-filter}
Figure 5.4 Filter your associations

If you open the diagram panel (Figure 5.5 ) and select an association in the table, you will get two graphic representations of this rule: the association diagram on the left indicating the actual intersection between the antecedent and consequent itemsets of this rule, and the intersection if unrelated diagram on the right, indicating the intersection if both itemsets were independent. These diagrams provide a visual overview of the importance of the selected association rule. The blue bar represents the portion of instances in the dataset that contain the antecedent items (Coverage), and the green bar is the portion of instances that contain the consequent itemset. The intersection between them is the portion of instances that contain both itemsets (i.e., the support of the rule). You can also get a visual insight of the lift and leverage rules, which are represented by the differences between both diagrams intersections, i.e., the differences of the actual association intersection versus the intersection if both itemsets were independent.

\includegraphics[]{images/table-diagram}
Figure 5.5 Association rule diagrams

Regarding the table (Figure 5.6 ), the main part of this view, each row contains a rule which is composed of two parts: the Antecedent itemset, with one or more items, and the Consequent itemset, which will always contain one item. For each rule you will find five different measures (Coverage, Support, Confidence, Leverage, and Lift) that describe the relationship between both parts of the rule. (The technicalities behind these rules are explained in section 2.1 .)

\includegraphics[]{images/table-table}
Figure 5.6 Associations discovered