TY - GEN
T1 - An iterative algorithm for sample selection based on the Reachable and Coverage
AU - WANG, Xizhao
AU - WU, Bo
AU - HE, Yulin
N1 - This research is partially supported by the Natural Science Foundation of Hebei Province (F2008000635), by the key project foundation of applied fundamental research of Hebei Province (08963522D), by the plan of 100 excellent innovative scientists of the first group in Education Department of Hebei Province, and by the Scientific Research Foundation of Hebei Province (06213548).
PY - 2009
Y1 - 2009
N2 - To overcome the drawbacks that Nearest Neighbour classification requires huge computation and memory storage, this paper proposes a new algorithm (ISSARC : Iterative Sample Selection Algorithm based on Reachable and Coverage) based on the conceptions of Reachable and Coverage. In this algorithm, a new function is introduced to evaluate the classification ability for each sample. According to the measuring function, a sample with the best classification ability is added to the subset and the samples which can be classified correctly are deleted in each iteration until the condensed subset is no longer getting smaller. It can be seen from analysis that time complexity of ISSARC is O (in2). The experimental results on two artificial data sets and some real data sets demonstrate the effectiveness and the feasibility of the proposed algorithm. Compared to traditional methods, such as MCS, ICF and ENN, the condensed sets obtained by ISSARC is superior in storage and classification accuracy.
AB - To overcome the drawbacks that Nearest Neighbour classification requires huge computation and memory storage, this paper proposes a new algorithm (ISSARC : Iterative Sample Selection Algorithm based on Reachable and Coverage) based on the conceptions of Reachable and Coverage. In this algorithm, a new function is introduced to evaluate the classification ability for each sample. According to the measuring function, a sample with the best classification ability is added to the subset and the samples which can be classified correctly are deleted in each iteration until the condensed subset is no longer getting smaller. It can be seen from analysis that time complexity of ISSARC is O (in2). The experimental results on two artificial data sets and some real data sets demonstrate the effectiveness and the feasibility of the proposed algorithm. Compared to traditional methods, such as MCS, ICF and ENN, the condensed sets obtained by ISSARC is superior in storage and classification accuracy.
KW - ENN
KW - ICF
KW - MCS
KW - Nearest neighbour rule
KW - Noise
KW - Sample selection
UR - http://www.scopus.com/inward/record.url?scp=74549219365&partnerID=8YFLogxK
U2 - 10.1109/ICCOMTA.2009.5349146
DO - 10.1109/ICCOMTA.2009.5349146
M3 - Conference paper (refereed)
AN - SCOPUS:74549219365
SN - 9781424448166
T3 - IEEE International Conference on Communications Technology and Applications, ICCTA
SP - 521
EP - 526
BT - Proceedings of 2009 IEEE International Conference on Communications Technology and Applications, IEEE ICCTA2009
PB - IEEE
T2 - 2009 IEEE International Conference on Communications Technology and Applications, IEEE ICCTA2009
Y2 - 16 October 2009 through 18 October 2009
ER -