8th INFORMS Workshop on Data Mining and Health Informatics
October 5, 2013
$120.00 Regular Registration and $50.00 Student Registration.
Important Dates for DM-HI Workshop
July 29: 200-word abstracts due
August 5: Acceptance decisions sent out
September 16: Proceedings papers due (final date to make changes)
October 5: DM-HI Workshop
Dorit S. Hochbaum
Professor of Industrial Engineering and Operations Research
University of California, Berkeley
An effective combinatorial optimization algorithm for data mining and clustering
We present here a novel methodology of data mining/machine learning that is based on combinatorial optimization. The technique considered takes into account pairwise relationships of similarity or dissimilarity between the data set's objects. The method is based on solving the optimization problem of "normalized cut prime" (NC'). This problem is closely related to the NP-hard problem of normalized cut, yet is polynomial time solvable.
NC' is a natural clustering problem involving two criteria. Given a collection of objects with pairwise similarity measure, the problem is to find a cluster that is as dissimilar as possible from the complement, while having as much similarity as possible within the cluster. The two objectives are combined with linear weights, selected (parametrically) for best trade-off. The problem, and its parametric version, is solved within the complexity of a single minimum s,t-cut algorithm.
The major features of NC' are reviewed and contrasted with those of the most commonly used data mining and machine learning methods, including Support Vector Machines (SVM), neural networks, PCA, logistic regression and others. The NC' algorithm is a non-geometric technique and unlike several other methods, it accepts the input either in the form of feature vectors or in any other format, provided that pairwise similarities are given for "neighbors" of each object. The technique can be utilized in supervised, or unsupervised, modes. An important feature is that it solves an optimization problem, which allows to link the results with the input considered.
We report on a specific application of NC' to the evaluation of drugs' effectiveness based on a data set in the form of cell images. In a comprehensive empirical study recently conducted on eleven data sets, in UCI data mining benchmark, the NC' technique was shown to be the leader in terms of accuracy.
Panos M. Pardalos
Distinguished Professor of Industrial and Systems Engineering
University of Florida
Data Mining and Knowledge Discovery in Dynamic Networks
In recent years, data mining and optimization heuristics have been used to analyze many large (and massive) data-sets that can be represented as a network. In these networks, certain attributes are associated with vertices and edges. This analysis often provides useful information about the internal structure of the datasets they represent. We are going to discuss our work on several networks from telecommunications (call graph), financial networks (market graph), social networks, and brain networks in neuroscience.
The Data Mining (DM), Artificial Intelligence (AI), and Health Applications (HA) Subdivisions of the Institute for Operations Research and Management Sciences (INFORMS) propose to jointly organize the 8th Pre-conference Workshop in conjunction with the 2013 INFORMS Annual Conference in Minneapolis, Minnesota. On behalf of the INFORMS DM-HI 2013 Organizing Committee, we would like to invite you to submit an abstract, not exceeding 200 words, for review. Authors of accepted abstracts will be expected to give a presentation at the Workshop and submit a short paper, not exceeding 6 pages, for the Workshop Proceedings, which will be available on a CD at the workshop.
The proposed workshop will consist of two parallel tracks, a data mining track and a health informatics track. Each track will consist of at least two morning sessions and two afternoon sessions, and each session will consist of three or four presentations.
Presentations and Proceedings
If you have been selected to give a presentation at the workshop, you are also expected to submit a short paper, not exceeding six pages, for the Workshop Proceedings that will be available on a CD at the workshop. Each presentation is 20 minutes long, with 5 minutes for questions.
Please ensure that the paper is checked carefully for grammatical problems, particularly if you are not a native English speaker.
– To register for the workshop, it is easiest to do this while you are registering for the INFORMS conference itself. The option to select the workshop is during the check-out.
– In order to expedite transitions between presentations at the workshop, we will try to have most of the presentations pre-loaded onto a laptop. Please either bring a PDF file with you on a USB drive, or email it to the organizers in advance.
For inquiries, please contact Program co-Chairs:
Onur Seref (Virginia Tech): email@example.com
Nicoleta Serban (Georgia Tech): firstname.lastname@example.org
Daniel Zeng (University of Arizona): email@example.com
Victoria Chen (University of Texas at Arlington)
Dolores Romero Morales (University of Oxford)
Kwok-Leung Tsui (City University of Hong Kong)
Helen Hao Zhang (University of Arizona)