Improved Boosting Algorithms Using Confidence-rated Predictions
Machine Learning - The Eleventh Annual Conference on computational Learning Theory
A family of additive online algorithms for category ranking
The Journal of Machine Learning Research
Learning multi-label alternating decision trees from texts and data
MLDM'03 Proceedings of the 3rd international conference on Machine learning and data mining in pattern recognition
LIBSVM: A library for support vector machines
ACM Transactions on Intelligent Systems and Technology (TIST)
Combining contents and citations for scientific document classification
AI'05 Proceedings of the 18th Australian Joint conference on Advances in Artificial Intelligence
| Hi-index | 0.00 |
This paper is devoted to the multi-label classification problem. We propose two new methods for reduction from ranking to multi-label case. In existing methods complex threshold functions are being defined on the input feature space, while in our approach we propose to construct simple linear multi-label classification functions on the class relevance space using the output of ranking algorithms as an input. In our first method we estimate the linear threshold function defined on the class relevance space. In the second method we define the linear operator mapping class ranks into the set of values of binary decision functions. In comparison to existing methods our methods are less computationally expensive and in the same time they demonstrate competitive and in some cases significantly better accuracy results in experiments on well-known multi-label benchmarks datasets.