Building Fuzzy OLAP Using Multi-attribute Summarization


D. P. V. Kasinadh and P. Radha Krishna
Institute for Development and Research in Banking Technology,
IDRBT, Castle Hills, Masab Tank, Hyderabad, INDIA.
dpvkasinadh@mtech.idrbt.ac.in, prkrishna@idrbt.ac.in

Abstract
Data Warehouse helps the decision makers of an organization in taking decisions that
helps in improving the profitability of business by consolidating and aggregating data from
many heterogeneous sources. Information available in this aggregated data is raw numbers.
These raw numbers does not provide semantics about the data to decision makers. For
example, “ A sale of amount 100000 is good or bad is unclear”. Usually the relationship
between the data and requirements to the decision maker are fuzzy in nature, rather than
crisp numbers. There is a need to design data-warehouse in such a way that it should address
the requirements of intelligent decision-making. In this paper, we build a fuzzy OLAP cube to
support qualitative data analysis by using multi-attribute summarization. Data is fuzzified
and assigned membership values using a cluster-based approach. To demonstrate the model,
we developed a prototype data warehouse for foreign exchange currency transactions and
analyzed these transactions with fuzzy OLAP operations.

1. Introduction
Data warehousing has gained prominence in the recent past because of their ability to
integrate data available across the enterprise from different sources. Usually data in a data
warehouse is stored in a multidimensional model. Design of the dimensional model for
building data warehouse and its advantages are presented by Kimball [6], which explains
about facts and dimensions. After integration, data warehouse contains aggregated and
summarized dimensions to facilitate querying and analysis in turn for decision-making and to
detect trends and anomalies [2]. These aggregations are done in order to reduce the cost
involved in calculating the aggregates for queries each time. On Line Analytical Processing
(OLAP) presents an approach to data analysis where data is consolidated and aggregated with
respect to multiple dimensions of interest in a data warehouse.
Agarwal, Gupta and Sarawagi [1] proposed a data model and a few algebraic operations
that provide semantic foundation to multidimensional databases on which OLAP is based.
This model provides support for multiple hierarchies along each dimension and support for
ad-hoc aggregates. OLAP tools usually provide operations such as roll-up, drill-down, slice
and dice on multidimensional model. The response out of these operations is usually in
quantitative values. In most cases, users require results in qualitative rather than quantitative
terms, as the relationship between data and queries on data are fuzzy in nature. For example, a
decision maker actually wants to know whether, the sale of a product in a particular location
is good, average or bad rather than crisp value as they may help greatly. The semantics rather
than raw numbers convey the real meaning of the data. If raw numbers are provided for
analysis, converting the quantitative value to qualitative may become a difficult process, as
multiple attributes may influence the decision.
International Conference on Computational Intelligence and Multimedia Applications 2007
0-7695-3050-8/07 $25.00 © 2007 IEEE
DOI 10.1109/ICCIMA.2007.201
370