Database Reference
In-Depth Information
One advantage of OLAP systems is that they can be evaluated using a
standardized set of benchmarks; for example, the OLAP Council APB-1
performance benchmark simulates a realistic OLAP business situation [4]. The
goal of APB-1 is to measure overall OLAP performance rather than the
performance of specific tasks. The operations performed during the APB-1 test
include: bulk loading of data, incremental loading of data, aggregation of data
along hierarchies, calculation of new data based on business models, time series
analysis, queries with a high degree of complexity, drill-down through hierarchies,
ad hoc queries, and multiple online sessions. In short, OLAP provides fast data
summarization and basic data processing. It can be used as one of the
preprocessing tools during the DMKD process, to make it more efficient and
easier to perform. Also, OLAP technology can be directly integrated with a
majority of other DM algorithms like association rules, classification, prediction,
and clustering [38].
OLAP is well coupled with DW because the data warehouses are designed
differently from traditional relational DBMS. DW is a central data repository that
defines integrated data models for data normally stored in a number of different
locations. It incorporates a subject-oriented read-only historical data. This not only
guarantees stability of the data but also gives flexibility to effectively query the
data stored in a warehouse.
1.3.7. OLE DB-DM
OLE DB-DM (OLE DB for Data Mining) is an extension of the SQL query
language that allows users to train and test DM models [51]. Its primary use is to
integrate different DM tools using a common API. The OLE DB-DM supports all
of the most popular DM tools and applies DM analysis directly against a relational
database. OLE DB-DM consists of these elements:
z a data mining model (DMM) is modeled by a relational table, except that it
contains columns used for training and predictions. After the data are inserted
into the table, a DM algorithm processes them and the resulting DM model is
saved. Thus, the DMM can be browsed, refined, or used.
z prediction join operation , an operation that does a join query between a
trained DM model and data to generate a prediction result that can be sent to
the user's application as either an OLE DB row set or an ADO (active data
objects) record set.
z OLE DB-DM schema row sets , which allow user applications to find available
DM services and models and the model contents.
One of the advantages of the OLE DB-DM is its support of standard DM data
types by using flags, instead of using only the SLQ data types. The following data
types are supported:
z key - discrete attribute that is a key.
z continuous - attribute with continuous values.
z discrete - attribute with discrete values.
z discretized - attribute is continuous and should be discretized.
Search WWH ::




Custom Search