Name
Abstract | ||
|---|---|---|
In this research we propose to derive new features based on data samples' local information with the aim of improving the performance of general supervised learning algorithms. The creation of new features is inspired by the measure of average precision which is known to be a robust measure that is insensitive to the number of retrieved items in information retrieval. We use the idea of average precision to weight the neighbours of an instance and show that this weighting strategy is insensitive to the number of neighbours in the locality. Information captured in the new features allows a general classifier to learn additional useful peripheral knowledge that are helpful in building effective classification models. We comprehensively evaluate our method on real datasets and the results show substantial improvements in the performance of classifiers including SVM, Bayesian networks, random forest, and C4.5. |
Year | DOI | Venue |
|---|---|---|
2014 | 10.1145/2661829.2662045 | CIKM |
Keywords | Field | DocType |
average precision,classification,data mining,local information | Data mining,Locality,Weighting,Pattern recognition,Computer science,Support vector machine,Bayesian network,Supervised training,Artificial intelligence,Classifier (linguistics),Random forest,Machine learning | Conference |
Citations | PageRank | References |
0 | 0.34 | 9 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Wei Liu | 1 | 468 | 37.36 |
| Dong Lee | 2 | 0 | 0.34 |
| Rao Kotagiri | 3 | 0 | 1.01 |