---

What: KSC is a data manipulation component that creates additional data describing the sequence of event transitions for analysis. This additional data can be very useful in understanding and predicting information from event data like customer behaviour. Like KEL, this component merges information from a static table with dynamic information from a transactional table. KSC aggregates by event, capturing the transitions in sequence between different events.

Why: Behavioural data can be much more predictive that purely demographic or static data. A customer may have had a number of events or transactions, such as purchase history, call centre complaints, claims, credit-card usage or web log data. These lists of events have different lengths for each customer (not all customers make the same number of claims). The sequence of events that led a customer to a certain stage can be a strong predictor for the next activity.

How: KSC creates aggregates of event transitions taken from a transactional table. It creates one line of data per customer from tables of events, generating sequence and transition information about customer behaviour. That way it is possible to create models predicting the next activity of a customer, for example, which product s/he is most likely to purchase.

Benefits for the business user: KSC does not require programming to perform this sophisticated aggregation. The user can quickly add valuable sequence data to the predictive modelling process.

Benefits for the Data Mining expert: KSC allows Data Mining professionals to incorporate transactional data in the modelling process. KSC is fast and can handle very large data sets.

Benefits for the Integration specialist and IT: Only two passes of the log table are required using an efficient internal data representation. Building transactional aggregates can be done in minutes instead of days.

Example: KSC can be very useful in analysing web log sessions. The reference table contains information about the sessions, and the transaction table contains the click stream. KSC is able to represent each session as a series of transitions between pages. The transition data makes it possible to predict when people are about to leave the site or when they should receive certain information during the session.

Copyright © 2002–2003 KXEN Inc. All rights reserved.