An Analysis of Order Dependence in k-NN
In classification based on k-NN with majority voting, the class assigned to a given problem is the one that occurs most frequently in the k most similar cases (or instances) in the dataset. However, different versions of k-NN may use different strategies to select the cases on which the solution is based when there are ties for the kth most similar case. One strategy is to break ties for the kth most similar case based on the ordering of cases in the dataset. We present an analysis of the order dependence introduced by this strategy and its effects on the algorithm’s performance.
Keywordsclassification k-NN instance-based learning case-based reasoning
Unable to display preview. Download preview PDF.
- 6.Brooks, A.D.: knnflex: A More Flexible KNN, http://cran.r-project.org/web/packages/knnflex
- 7.R Development Core Team: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (2009)Google Scholar
- 9.Langley, P.: Order Effects in Incremental Learning. In: Reimann, P., Spada, H. (eds.) Learning in Humans and Machines: Towards an Interdisciplinary Learning Science. Elsevier, Oxford (1995)Google Scholar
- 13.Asuncion, A., Newman, D.J.: UCI Machine Learning Repository. University of California, Irvine, School of Information and Computer Sciences (2007)Google Scholar
- 14.Kohavi, R.: A Study of Cross-Validation and Bootstrap for Accuracy Estimation and Model Selection. In: 14th International Joint Conference on Artificial Intelligence, pp. 1137–1143. Morgan Kaufmann, San Mateo (1995)Google Scholar