Artificial endocrine system and its application for nearest neighbor rule condensation

作者：Wang Zhanmin^*; Cui Duwu; Xu Qingzheng

来源：Information Technology Journal, 2011, 10(11): 2004-2013.

DOI：10.3923/itj.2011.2004.2013

• Computational costs
• Consistent subset
• Data sets
• Endocrine systems
• Nearest neighbor classification
• Nearest neighbor rule
• Original sample
• Prototype reduction algorithms
• Reference set
• Rule set
• Training sets

Prototype reduction algorithms are very useful in the area of nearest neighbor classification and most of them process the given datasets in its entirety. The main disadvantage of them is the excessive computational cost especially when the prototype size is very large. A divide-reduce-coalesce mechanism has been proposed to overcome this shortcoming. But, the final subsets obtained with most of the traditional algorithms in this divide-reduce-coalesce manner are not consistent subsets of original sample set. So, the accuracies of the classifiers obtained in the manner are not been guaranteed too. In this study, we present an artificial endocrine system to reduce useless points in a given training set. The method remains only the points on the boundary between different classes and the amount of reduced rules of the reference set can be revised by the granularity of the lattice. The main advantages of the proposed method are, on the one hand that it is able to condense a given rule set within less time comparing with other traditional algorithms. On the other hand, with the proposed algorithm we can get a consistent subset of the given set in the divide-reduce-coalesce manner. This approach has been tested using 13 different datasets. The experiments show successful results when the size of the given dataset is large.

• 出版日期2011