원문정보
초록
영어
To overcome the drawback of KNN algorithms based on distance measure which did not measure the contributions for each feature accurately. In this paper, a K-Nearest Neighbor (KNN) de-noise method based on likelihood distance entropy is proposed. The relations of feature parameters are used to measure their contributions for de-noise energy, then according to the contributions for each feature leading de-noise of the feature parameters. In order to compare the performance of these relative methods, the Letter corpora and Pima Indians Diabetes data-base are employ to carry out the experiments, the experiment results show that comparing with the other de-noise methods mentioned in this paper, this proposed method have a better ability for de-noise.
목차
1. Introduction
2. Theoretical Background
2.1. Similarity Distance Noise Reduction Entropy
2.2. Implementation
2.3. Accuracy Analysis
3. Experiment Setup and Analysis
3.1. Experiment Setup
3.2. Result Analysis
4. Conclusion
Acknowledgements
References