Syntax of KDSL for specifying different kinds of knowledge

KDSL adopts an SQL­like syntax to facilitate high level data mining and natural integration with relational query language, SQL.

The KDSL language is defined in an extended BNF grammar, where "[ ]" represents 0 or one occurrence, "{}" represents 0 or more occurrences, as shown on the next page.

 

Figure: Architecture of Intension Mining System

<KDSL> : : =

Schema_Name is <schema_name>

Accumulate <knowledge_type>

Use database <database_name>

{Use hierarchy <hierarchy_name> for <attribute>}

related to <attr_list>

from <relation>

[use selection <selection_function>]

[use cleaning <cleaning_function>]

[use transformation <transformation_function>]

use mining <mining_algorithm>

{with [<kind_of>] parameter = <parameter_value>}

In <KDSL>, " Schema_Name is <schema_name>" is to specify the name of the schema. It can be any valid variable name.

The statement "Accumulate <knowledge_type>", is the specification of the kind of knowledge to be discovered. It can be either classification, clustering or association rules.

The statement "Use database <database_name>" directs the mining task to a specific database "<database name>", and the optional statement, "Use hierarchy <hierarchy_name> for <attribute>", assigns <hierarchy> to a particular attribute <attribute>.

The related­to statement, "related to <attr_list>", selects a list of relevant attributes for mining. The "from <relation>" statement specifies the relevant relation to be used from the database <database_name>.

The optional statements "use selection <selection_function>", "use cleaning <cleaning_function>" and "use transformation <transformation_function>" specify the selection function, cleaning function and the transformation function if at all required by the user.

The "use mining <mining_algorithm>" is used to specify the algorithm used to mine the knowledge. It should be in correspondence with the <knowledge_type> we are discovering.

The "with­parameter" statement specifies various kinds of parameters like priority, periodicity, start time, etc and the knowledge dependent parameters like confidence, support, no. of clusters, etc.

Examples of mining different kinds of knowledge

Let us examine a university database with the following schema.

student(name, sno, status, major, cgpa, date_of_birth, address)

course(cno, title, department)

A few examples of knowledge discovery schema defined using KDSL are presented as follows.

 

1. Mining association rules.

To find all the association rules of the graduate students in relevance to the attributes cgpa, date of birth and address.

Schema_Name is kds01

accumulate association rules

use database university_database

related to cgpa date_of_birth address

from student

with priority parameter = 1

with periodicity parameter = 2000(seconds)

with start_time parameter = 0700(hours)

with antecedent parameter = *

with consequent parameter = *

 

2. Mining clusters.

To find 10 clusters in relevance to attributes cno and title.

Schema_Name is kds02

accumulate clustering rules

use database university_database

related to cno title

from course

with priority parameter = 2

with periodicity parameter = 3600(seconds)

with start_time parameter = 1100(hours)

with k_means parameter = 10