Multi-label weighted k-nearest neighbor classifier with adaptive weight estimation

Authors:
Jianhua Xu
Affiliations:
School of Computer Science and Technology, Nanjing Normal University, Nanjing, China
Venue:
ICONIP'11 Proceedings of the 18th international conference on Neural Information Processing - Volume Part II
Year:
2011

Citing 10
Cited 0

A Comparison of Ranking Methods for Classification Algorithm Selection

ECML '00 Proceedings of the 11th European Conference on Machine Learning
Nonparametric Supervised Learning by Linear Interpolation with Maximum Entropy

IEEE Transactions on Pattern Analysis and Machine Intelligence
Multilabel Neural Networks with Applications to Functional Genomics and Text Categorization

IEEE Transactions on Knowledge and Data Engineering
ML-KNN: A lazy learning approach to multi-label learning

Pattern Recognition
Top 10 algorithms in data mining

Knowledge and Information Systems
On kernel difference-weighted k-nearest neighbor classification

Pattern Analysis & Applications - Special Issue: Non-parametric distance-based classification techniques and their applications
Ml-rbf: RBF Neural Networks for Multi-Label Learning

Neural Processing Letters
Feature selection for multi-label naive Bayes classification

Information Sciences: an International Journal
Similarity-based Classification: Concepts and Algorithms

The Journal of Machine Learning Research
Combining instance-based learning and logistic regression for multilabel classification

Machine Learning

Quantified Score

Hi-index	0.00

Visualization

Abstract

Multi-label classification is a special learning task in which any instance is possibly associated with multiple classes simultaneously. How to design and implement efficient and effective multi-label algorithms is a challenging issue. The k-nearest neighbor (kNN) method and its weighted form (WkNN) are simple but effective for binary and multi-class classification. In this paper, we construct a weighted kNN version for multi-label classification (MLC-WkNN) according to Bayesian theorem. Through approximating a query instance by the linear weighted sum of k-nearest neighbors in terms of least squared error, the weights are determined adaptively by solving a quadratic programming with a unit simplex constraint. Specially, our MLC-WkNN is still a model-free and instance-based learning technique and only involves a tunable parameter k. Experimental study on two benchmark data sets (Image and Yeast) illustrates that our MLC-WkNN outperforms seven existing high-performed multi-label algorithms.