Optimizing the Area Under a Receiver Operating Characteristic Curve With Application to Landmine Detection

A common approach to training neural network classifiers in a supervised learning setting is to minimize the mean-square error (mse) between the network output for each labeled training sample and some desired output. In the context of landmine detection and discrimination, although the performance of an algorithm is correlated with the mse, it is ultimately evaluated by using receiver operating characteristic (ROC) curves. In general, the larger the area under the ROC curve (AUC), the better. We present a new method for maximizing the AUC. Desirable properties of the proposed algorithm are derived and discussed that differentiate it from previously proposed algorithms. A hypothesis test is used to compare the proposed algorithm to an existing algorithm. The false alarm rate achieved by the proposed algorithm is found to be less than that of the existing algorithm with 95% confidence

[1]  J. Kacprzyk,et al.  The Ordered Weighted Averaging Operators: Theory and Applications , 1997 .

[2]  Nirmal K. Bose,et al.  Landmine detection and classification with complex-valued hybrid neural network using scattering parameters dataset , 2005, IEEE Transactions on Neural Networks.

[3]  Michael C. Mozer,et al.  Optimizing Classifier Performance via an Approximation to the Wilcoxon-Mann-Whitney Statistic , 2003, ICML.

[4]  Ronald R. Yager,et al.  On ordered weighted averaging aggregation operators in multicriteria decisionmaking , 1988, IEEE Trans. Syst. Man Cybern..

[5]  Paul D. Gader,et al.  Landmine detection using fuzzy sets with GPR images , 1998, 1998 IEEE International Conference on Fuzzy Systems Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98CH36228).

[6]  Paul D. Gader,et al.  Landmine detection with ground penetrating radar using fuzzy k-nearest neighbors , 2004, 2004 IEEE International Conference on Fuzzy Systems (IEEE Cat. No.04CH37542).

[7]  Alain Rakotomamonjy,et al.  Optimizing Area Under Roc Curve with SVMs , 2004, ROCAI.

[8]  Paul D. Gader,et al.  Detection of land mines using fuzzy and possibilistic membership functions , 2003, The 12th IEEE International Conference on Fuzzy Systems, 2003. FUZZ '03..

[9]  János C. Fodor,et al.  Characterization of the ordered weighted averaging operators , 1995, IEEE Trans. Fuzzy Syst..

[10]  Joseph N. Wilson,et al.  Detecting landmines with ground-penetrating radar using feature-based rules, order statistics, and adaptive whitening , 2004, IEEE Transactions on Geoscience and Remote Sensing.

[11]  Paul D. Gader,et al.  Continuous Choquet integrals with respect to random sets with applications to landmine detection , 2004, 2004 IEEE International Conference on Fuzzy Systems (IEEE Cat. No.04CH37542).

[12]  Jing Zhang,et al.  Landmine Feature Extraction and Classification of GPR Data Based on SVM Method , 2004, ISNN.

[13]  Otto Lohlein,et al.  Classification of GPR data for mine detection based on hidden Markov models , 1998 .

[14]  Janusz Kacprzyk,et al.  The Ordered Weighted Averaging Operators , 1997 .

[15]  Chandra S. Throckmorton,et al.  Algorithms for land mine detection using the NIITEK ground penetrating radar , 2002, SPIE Defense + Commercial Sensing.

[16]  Mahmood R. Azimi-Sadjadi,et al.  Structural adaptation in neural networks with application to land mine detection , 1997, Defense, Security, and Sensing.

[17]  Ping Chen,et al.  Training DHMMs of mine and clutter to minimize landmine detection errors , 2003, IEEE Trans. Geosci. Remote. Sens..

[18]  Mahmood R. Azimi-Sadjadi,et al.  Detection of mines and minelike targets using principal component and neural-network methods , 1998, IEEE Trans. Neural Networks.

[19]  Ronald R. Yager,et al.  On ordered weighted averaging aggregation operators in multicriteria decision-making , 1988 .

[20]  Paul D. Gader,et al.  Landmine detection with ground penetrating radar using hidden Markov models , 2001, IEEE Trans. Geosci. Remote. Sens..

[21]  Leslie M. Collins,et al.  Application of texture feature classification methods to landmine/clutter discrimination in off-lane GPR data , 2004, IGARSS 2004. 2004 IEEE International Geoscience and Remote Sensing Symposium.

[22]  Paul D. Gader,et al.  Multi-sensor and algorithm fusion with the Choquet integral: applications to landmine detection , 2004, IGARSS 2004. 2004 IEEE International Geoscience and Remote Sensing Symposium.