Multiple classifier system in class-imbalanced problems and large data sets

• Grant number: 11/16411-4
• Support Opportunities: Regular Research Grants
• Start date: December 01, 2011
• End date: November 30, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Moacir Antonelli Ponti
• Grantee: Moacir Antonelli Ponti
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Classification methods often fail to address two common problems in real applications: large data sets and imbalanced class data. Data sets with large number of elements are becoming more pervasive due to the availability of sensors and storage technology, and by the nature of some applications such as financial transactions, network logs and bioinformatics. Multiple classifier systems can be used both to paralelize or distribute the processing, and to allow the undersampling of the training set, so that large data sets are feasible to be used in classification tasks. Methods derived from Bagging techniques can be used to generate diversity in multiple classifiers, trained in parallel with small amount of data, resulting in similar or higher accuracies when compared to a simple classification. The use of ensembles of classifiers also have the potential to minimize the class-imbalance problem, by sampling methods applied with Boosting techniques, aided by concept drift studies. This projects aims the investigation of these two issues by the point of view of multiple classifier systems, researching solutions with applications in various areas. (AU)

Scientific publications

(4)

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

PONTI, MOACIR; NAZARE, TIAGO S.; THUME, GABRIELA S.. Image quantization as a dimensionality reduction procedure in color and texture feature extraction. Neurocomputing, v. 173, n. 2, SI, p. 385-396, JAN 15 2016. (11/22749-8, 11/16411-4)

PONTI, MOACIR; NAZARE, TIAGO S.; THUME, GABRIELA S.. Image quantization as a dimensionality reduction procedure in color and texture feature extraction. Neurocomputing, v. 173, p. 12-pg., 2016-01-15. (11/22749-8, 11/16411-4)

PONTI, JR., MOACIR P.. Segmentation of Low-Cost Remote Sensing Images Combining Vegetation Indices and Mean Shift. IEEE Geoscience and Remote Sensing Letters, v. 10, n. 1, p. 67-70, JAN 2013. (11/16411-4)

COSTA, GABRIEL B. P.; PONTI, MOACIR; FRERY, ALEJANDRO C.; ZHOU, ZH; SCHWENKER, F. Partially Supervised Anomaly Detection Using Convex Hulls on a 2D Parameter Space. PARTIALLY SUPERVISED LEARNING, PSL 2013, v. 8193, p. 8-pg., 2013-01-01. (11/16411-4, 12/12524-1)