Mining Your Own Business in Telecoms Using DB2 Intelligent Miner for Data

Chapter 8. The value of DB2 Intelligent Miner for Data 153

Value prediction

Value prediction is similar to classification; the goal is to build a data model as a

generalization of the records. However, the difference is that the target is not a

class membership but a continuous value, or ranking. IM for Data has two

prediction algorithms: a neural network algorithm and a Radial Basis Functions

(RBF) algorithm. The radial basis function is particularly efficient and is

appropriate for value prediction with very large data sets.

Similar time sequences

The purpose of this process is to discover all occurrences of similar

subsequences in a database of time sequences. Given a database of time

sequences, the goal is to find sequences similar to a given one, or find all

occurrences of similar sequences. The powerful alternatives afforded by multiple

methods are enhanced by the fact that several of the methods are supported by

more than one mining technique. Multiple techniques are often used in

combination to address a specific business problem.

8.2.4 Creating and visualizing the results

Information that has been created using statistical or mining functions can be

saved for further analysis in the form of result objects. The result objects can be

visualized using a variety of graphical displays or the results exported to

spreadsheets (for example, EXCEL, LOTUS 123), or to browsers (for example,

Netscape, Explorer), or to specific statistical packages (for example, SPSS).

Result objects can be used in several ways:

򐂰 To visualize or access the results of a mining or statistical function

򐂰 To determine what resulting information you want to write to an output data

object

򐂰 To be used as input data, when running a mining function in test mode to

validate the predictive model representation by the result

򐂰 To be used as input data, when running a mining function in application mode

to apply the model to new data

8.3 DB2 Intelligent Miner Scoring

DB2 Intelligent Miner Scoring (IM Scoring) is an economical and easy-to-use

mining deployment capability. It enables users to incorporate analytic mining into

Business Intelligence, eCommerce and OLTP applications. Applications score

records (segment, classify or rank the subject of those records) based on a set of

predetermined criteria expressed in a data mining model.

154 Mining Your Own Business in Telecoms Using DB2 Intelligent Miner for Data

These applications can better serve business and consumer users alike — to

provide more informed recommendations, to alter a process based on past

behavior, to build more efficiencies into the online experience; to, in general, be

more responsive to the specific situation at hand. All scoring functions offered by

the DB2 Intelligent Miner for Data are supported.

The IM Scoring is an add-on service to DB2, consisting of a set of User Defined

Types (UDTs) and User Defined Functions (UDFs), which extends the

capabilities of DB2 to include some data mining functions. Mining models

continue to be built using the IM for Data, but the mining application mode

functions are integrated into DB2. Using the IM Scoring UDFs, you can import

certain types of mining models into a DB2 table and apply the models to data

within DB2. The results of applying the model are referred to as scoring results

and differ in content according to the type of model applied. The IM Scoring

includes UDFs to retrieve the values of scoring results.

The results of applying the model are referred to as scoring results and differ in

content according to the type of model applied. The IM Scoring includes

functions to retrieve the values of scoring results.

The IM Scoring is available on the following operating systems:

򐂰 AIX

򐂰 Solaris

򐂰 Windows NT, Windows 2000

򐂰 Linux, Linux/390

Summary of functionality

The application mode for the following IM for Data mining and statistical functions

are supported by the IM Scoring:

򐂰 Demographic and neural clustering

򐂰 Tree and neural classification

򐂰 RBF and neural prediction

򐂰 Polynomial regression

Scoring functions are provided to work with each of these types. Each scoring

function includes different algorithms to deal with the different mining functions

included within a type, for example, the clustering type includes demographic and

neural clustering and so, scoring functions for clustering include algorithms for

demographic and neural clustering. For all the supported mining functions, you

build and store the model using the IM for Data. Models must then be exported to

an external file.