Abstract The traditional prototype selection algorithms are susceptible to pattern reading sequence, abnormal patterns etc. Aiming at these problems, an improved prototype selection algorithm based on adaptive boundary approximation is proposed by a detailed analysis of the prototype learning rule. The prototype absorption strategy of condensed nearest neighbor algorithm (CNN) is improved and the closer homogeneous boundary prototype parallel to its current nearest one is retained. Meanwhile, the prototype updating strategy is built for achieving dynamic periodic updating to the prototype set. The proposed algorithm can overcome the above mentioned issues and effectively reduce the scale of prototype set. Experiments are made on the artificial dataset and UCI benchmark dataset, and the results show that the final prototype set obtained by the proposed algorithm reflects the distribution of the original dataset much better. It improves the average reduction ratio performance, has better classification accuracy and runs faster than other algorithms.
[1] Altincay H. Improving the k-Nearest Neighbour Rule: Using Geometrical Neighbourhoods and Manifold-Based Metrics. Experts Systems, 2011, 28(4): 391-406 [2] Olvera-López J A. Prototype Selection Methods. Computacióny Sistemas, 2010, 13(4): 449-462 [3] Triguero I, Derrac J, Garcia S, et al. A Taxonomy and Experimental Study on Prototype Generation for Nearest Neighbor Classification. IEEE Trans on Systems, Man, and Cybernetics: Part C, 2011, 42(1): 86-100 [4] Czarnowski I. Cluster-Based Instance Selection for Machine Classification. Knowledge and Information Systems, 2012, 30(1): 113-133 [5] Abroudi A, Farokhi F. Prototype Selection for Training Artificial Neural Networks Based on Fast Condensed Nearest Neighbor Rule // Proc of the IEEE Conference on Open Systems. Kuala Lumpur, Malaysia, 2012. DOI: 10.1109/ICOS.2012.6417625 [6] Chang F, Lin C C, Lu C J. Adaptive Prototype Learning Algorithms: Theoretical and Experimental Studies. Journal of Machine Learning Research, 2006, 7: 2125-2148 [7] Gowda K, Krishna G. The Condensed Nearest Neighbor Rule Using the Concept of Mutual Nearest Neighborhood. IEEE Trans on Information Theary, 1979, 25(4): 488-490 [8] Sáez J A, Luengo J, Herrera F. Predicting Noise Filtering Efficacy with Data Complexity Measures for Nearest Neighbor Classification. Pattern Recognition, 2013, 46(1): 355-364 [9] Olvera-López J A, Carrasco-Ochoa J A, Martínez-Trinidad J F. A New Fast Prototype Selection Method Based on Clustering. Pattern Analysis and Applications, 2010, 13(2): 131-141 [10] Raicharoen T, Lursinsap C. A Divide-and-Conquer Approach to the Pairwise Opposite Class-Nearest Neighbor (POC-NN) Algorithm. Pattern Recognition Letters, 2005, 26(10): 1554-1567 [11] Fayed H A, Atiya A F. A Novel Template Reduction Approach for the K-Nearest Neighbor Method. IEEE Trans on Neural Networks, 2009, 20(5): 890-896 [12] Li S Z, Lu J W. Face Recognition Using the Nearest Feature Line Method. IEEE Trans on Neural Networks, 1999, 10(2): 439-443 [13] Pang Y W, Yuan Y, Li X L. Iterative Subspace Analysis Based on Feature Line Distance. IEEE Trans on Image Processing, 2009, 18(4): 903-907 [14] Qing J J, Huo H, Fang T. Pattern Classification Based on k Loca-lly Constrained Line. Soft Computing, 2010, 15(4): 703-712 [15] Gao Q B, Wang Z Z. Center-Based Nearest Neighbor Classifier. Pattern Recognition, 2007, 40(1): 346-349 [16] Mi J X, Huang D S, Wang B, et al. The Nearest-Farthest Subspace Classification for Face Recognition. Neurocomputing, 2013, 113: 241-250 [17] Du H, Chen Y Q. Rectified Nearest Feature Line Segment for Pa-ttern Classification. Pattern Recognition, 2007, 40(5):1486-1497 [18] Chen Y N, Han C C, Wang C T, et al. Face Recognition Using Nearest Feature Space Embedding. IEEE Trans on Pattern Analysis and Machine Intelligence, 2011, 33(6): 1073-1086 [19] Xu Y, Shen F R, Zhao J X. An Incremental Learning Vector Quantization Algorithm for Pattern Classification. Neural Computing and Applications, 2012, 21(6): 1205-1215
2015, Vol. 28 
