|
As data
collection channels and means become more and diverse, many real-world data
mining tasks can easily acquire multiple data sets from various information
sources. Compared to single-source mining problems in which all the data
for a mining task are in the same pattern representation and are assumed to
be drawn from the identical distribution, a multi-source mining problem is
built on multiple information sources which have different contributions to
the target task and can complement one another to boost the performance. To
better leverage multiple information sources, integrating and transferring
knowledge among multiple data sets has become a crucial step in data
mining.
We call
for papers for addressing data mining problems in multi-source scenarios.
On one hand, many data mining and analysis tasks can significantly improve
their performance if knowledge mined from multiple sources can be properly
integrated and shared. On the other hand, comparing patterns from different
data sources and understanding their relatedness can be beneficial for
applications ranging from social science to bioinformatics to economics.
Thus, it becomes urgent to develop theories, methods, applications, and
knowledge representations, for mining from multiple information sources
that share relatedness.
Topics of Interest
Representative issues to be addressed
include but are not limited to:
1.
Transfer learning from multiple
information sources
·
Transfer
learning from heterogeneous and structured data sources
·
Transfer
learning from stream data sources
·
Foundation
and theories of transfer learning and domain adaptation
2.
Pattern correlation and differentiation in
different data sources
·
Pattern
comparison across multiple data sources
·
Pattern
fusion and synthesizing from multiple data sources
·
Pattern
summarization from multiple data sources
3.
Integrative and cooperative mining
·
Model
integration and fusion from multiple information sources
·
Ensemble
learning from multiple data sources
·
Multi-view
learning from multiple data sources
4.
Data integration and harnessing complex
data relationship
·
Database
similarity assessment and quantification
·
Automatic
schema mapping and relationship discovery
·
New
mapping framework for multiple information sources
5.
Multi-source data mining applications and
case studies
·
Web and
social media mining
·
Reality
mining, urban and environment sensing
·
Bioinformatics
and biomedical data mining
Paper Submission
All
papers should be formatted to IEEE Computer Society
Proceedings Manuscript Formatting Guidelines with a maximum of 8 pages in the 2-column format.
Please visit the IEEE ICDM
2011 website for detailed formatting and submission guidelines. Papers
that do not comply with the Submission Guidelines will be rejected without
review. The workshop proceedings will be published by the IEEE Digital
Library.
|
