Demonstration – KL-Miner

# Example

In this part we will show an example of KLTask. We will use data of fictious bank Barbora where are 6 181 objects. For our analysis we will use these attributes devided into partial cedents:

Client with two attributes:

• Age – with 5 categories (intervals with step 10 years),
• Sex – with 2 categories (male, female).

Social Status with two attributes:

• Salary – with 3 categories,
• District – with 77 categories.

Row Attribute with one attribute:

• Duration – length of the loan with 5 categories.

Column Attribute with one attribute:

• Amount – the amount of borrowed money in thousands of Czech crowns devided into 6 categories (intervals).

Figure 1: Input parameters

We solve the task Is there any correlation between duration of the loan and amount of borrowed money? We can write this task as an association rule in the form:

Duration × Amount / Client Social Status

Figure 1 shows our task in general. Here can be seen that as quantifier is used Sum of values with value greater than 50 and Kendall´s coefficient with value greater than 0.750. Antecedent and succedent consist of one partial cedents each (Row Attribute, Column Attribute). Condition consists of two partial cedents (Client, Social Status). These partial cedents for condition can be seen in detail on next two figures. From each figure can be seen all necessary parameters as length of partial cedents and all parameters for each attribute.

Figure 2: Partial condition – Client

Figure 3: Partial condition – Social Status

After generating this task we get 24 hypothesis (see fig. 4) and one of them is displayed on fig. 5. This hypothesis says that there is high positive correlation of amount of borrowed money and duration of the loan between women with average salary at the age of 61 to 71.

Figure 4: All Hypotheses

Figure 5: Hypothesis detail