My Present research
interests are in supervised and unsupervised machine learning technique and
data mining algorithms and their contributions to several dimensionality
reduction methods based on low-rank approximations, with a focus on their combination
possibilities, their Classification (e.g. Logistics Regression, Random Forest,
SVM etc.), their simplification (e.g., initialization of NMF
algorithms), their interpretation (e.g., different variants of computing
low-rank approxima-tions), their parallelization/distribution (e.g.,
multi-core, multi-processor and computer cluster archi-tecture) and their high-performance
computing aspects.


Research Background

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!

order now


In the last years my
research activities focused on machine learning and data mining applications in
various fields. Besides text-mining and classification problems, I have worked
on classification algorithms and feature selection strategies in the field of Market
Research, Software Development and Health Care.


In the “Outlier analysis”
project, I developed and analyzed new methods for detection, defense and
prevention of unwanted mass claims (Fraud Claims). My particular research
focused at pre-authorized, Reimbursement Claims and Underwriting. I developed a
component-based architecture for claims filtering based on greylisting in
combination with a multi-level reputation system in order to provide a reliable
and very scalable, fraudulent claims filtering system. Another goal was to
develop methods for distinguishing between mostly soft fraud and potentially
“Fraud” or malicious phishing claim.


Second project, which I did,
is “Simulation of Glucose-Insulin Response Model” which focused on the
development of innovative methods and computational technologies and on their
application in selected normal and impaired patient’s data to model and simulate
problems in Glucose appearance and Insulin secretion. I worked on the
application of sim-biology for the parameter estimation using sbionlinfit and
sbioparamestim with a forcing function strategy.


In another project called “Building Profit Function model based on transaction Data” I
worked on survival, loyalty models
for computing market scenarios,
especially on customer purchase prediction. The goal of this project was to
develop methods and architectures for proactive data replication management for
shopping Mall, based
on brand survival,
loyalty models and on the prediction of future purchase behavior of customer.



In the last years my
research activities focused on various strategies for reducing the attribute
space within predictive machine learning techniques. Feature selection and
dimensionality reduction techniques provide means to reduce the dimension of
data in order to reduce the noise and redundancy, and to increase the
classification performance of machine learning algorithms. Most of my actual
research activities contribute to my PhD thesis which focuses on simplifying,
improving and distributing feature selection and dimensionality reduction
strategies for data mining applications. An important task in this context is
to analyze relationships between several feature selection and dimensionality
reduction methods and the resulting classification accuracy.


Currently, I am working in
the Predicting a chance of insurer will claim and it will be an outlier-project,
on initialization techniques for non-negative matrix factorization (NMF) as
well as parallelization, distribution and high-performance computing aspects of
low-rank approximation techniques (NMF, SVD, PCA, etc). Initialization
strategies can improve the NMF in terms of faster convergence and faster error
reduction, since the number of complete repetitions and iterations within
single NMF factorizations can be reduced significantly. Although the benefits
of good NMF initialization techniques are well known in the literature, rather
few NMF initializations methods have been published so far. Another part of my
current research is based on investigating and comparing of various algorithmic
variants for efficiently computing the NMF on multi-core systems.



Research Agenda


In the next years I plan to
continue my efforts in the broad field of data mining and machine learning
research. On the one hand, I want to focus on technical solutions to
efficiently utilize the parallelization and distribution capabilities of
dimensionality reduction methods and learning algorithms. These
high-performance computing aspects play a pivotal role for many data mining
applications since the amount of data has increased enormously over the last
years and will continue to do so. On the other hand, I am very interested in
interdisciplinary collaborations and to bring in my knowledge in data mining
and machine learning research to solve problems in other disciplines. I believe
that co-operations with other researchers from different fields will provide
interesting opportunities for both sides. This synergy of technical and
interdisciplinary developments allows for several important and interesting
directions for future work.


In summary, data mining and machine learning are exciting and active research
areas that offer solutions to many challenges spanning the areas of web and
media search, data warehousing, mass surveillance, costumer analytics, analysis
of on-line communities, pharmaco- and bioinformatics, etc. I believe that my
experience puts me in a strong position to tackle some of the challenges
related to my research area.

Post Author: admin


I'm Irvin!

Would you like to get a custom essay? How about receiving a customized one?

Check it out