doi:10.1016/j.inffus.2006.02.002 
Copyright © 2006 Published by Elsevier B.V.
An optimized architecture for classification combining data fusion and data-mining
George Giglia, Éloi Bossé
, a, 
and George A. Lampropoulosa
aDRDC Valcartier, 2459, Boulevard PIE XI North, Val Belair, Que., Canada G3J 1X5
Received 25 November 2004;
revised 14 February 2006;
accepted 22 February 2006.
Available online 5 May 2006.
References and further reading may be available for this article. To view references and further reading you must
purchase this article.
Abstract
This paper presents a new architecture to integrate a library of feature extraction, Data-mining, and fusion techniques to automatically and optimally configure a classification solution for a given labeled set of training patterns. The most expensive and scarce resource in any detection problem (feature selection/classification) tends to be the acquiring of labeled training patterns from which to design the system. The objective of this paper is to present a new Data-mining architecture that will include conventional Data-mining algorithms, feature selection methods and algorithmic fusion techniques to best exploit the set of labeled training patterns so as to improve the design of the overall classification system. The paper describes how feature selection and Data-mining algorithms are combined through a Genetic Algorithm, using single source data, and how multi-source data are combined through several best-suited fusion techniques by employing a Genetic Algorithm for optimal fusion. A simplified version of the overall system is tested on the detection of volcanoes in the Magellan SAR database of Venus.
Keywords: Data-mining; Feature selection; Data fusion; Classification
Fig. 1. Fusion Matrix, FM(o).
Fig. 2. Generation of the Data-mining Architecture in the case of a single source.
Fig. 3. Generation of the Data-mining Architecture in the case of multiple sources.
Fig. 4. (a) Experiment A1 of Hom4; (b) Experiment A2 of Hom4.
Fig. 5. (a) Experiment A2 of Hom4; (b) Experiment A3 of Hom4.
Fig. 6. Raw data for training samples of Experiment A.
Fig. 7. SVD for training samples of Experiment A.
Fig. 8. (a) Experiment A1 of Hom4 using 6 SVD features; (b) Experiment A2 of Hom4 using 6 NSVD features.
Fig. 9. (a) Experiment A3 of Hom4 using 6 NSVD features; (b) Experiment A4 of Hom4 using 6 SVD features.
Table 1.
Decision Tree configuration
Table 2.
Exp. A1 Bayesian Classifier
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition.
Table 3.
Exp. A2 Bayesian Classifier
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition.
Table 4.
Exp. A3 Bayesian Classifier
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition.
Table 5.
Exp. A4 Bayesian Classifier
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition.
Table 6.
Exp. A1 Decision Tree
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition; RAW, 225 raw data points.
Table 7.
Exp. A2 Decision Tree
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition; RAW, 225 raw data points.
Table 8.
Exp. A3 Decision Tree
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition; RAW, 225 raw data points.
Table 9.
Exp. A4 Decision Tree
TD, total detections; AD, actual detections; FA, false alarms; MD, missed detections; SVD, 6 elements of Singular Value Decomposition; NSVD, 6 elements of Normalized Singular Value Decomposition; RAW, 225 raw data points.
Table 11.
HOM4: combination of Exp. A1–A4 using Bayesian Classifier
Table 10.
Misclassification Error for each training set of HOM4
Table 12.
HOM4: combination of Exp. A1–A4 using Decision Trees
Abstract
Purchase PDF (1186 K)
E-mail Article
Add to my Quick Links
Cited By in Scopus (0)
Purchase PDF (594 K)
