SF服务编织>> FeatureSelectionWithJava>> 返回
项目作者： ben-ix

项目描述 ：
Common feature selection algorithms implemented in Java
高级语言： Java
项目地址: git://github.com/ben-ix/FeatureSelectionWithJava.git
创建时间： 2017-04-13T10:31:13Z
项目社区：https://github.com/ben-ix/FeatureSelectionWithJava
开源协议：
下载


Feature Selection

Common feature selection algorithms implemented in Java, including

This uses a Wrapper approach, utilising the Weka library as a classifier.

To see how to use view the TestAll class, which guives an example of creating a new FeatureSelection
object and shows how to utilise the various stopping methods.

Datasets used for testing:

Sample output from WINE dataset on random training:testing (70:30) split of the data

  1. -------------------
  2. Sequential backward floating selection for 10 features
  3. Accuracy using all features: 47.059%
  4. Accuracy using features ([2, 3, 6, 7, 8, 9, 10, 11]): 94.118%
  5. -------------------
  6. -------------------
  7. Sequential backward floating selection
  8. Accuracy using all features: 41.176%
  9. Accuracy using features ([0, 1, 2, 8, 9, 11]): 88.235%
  10. -------------------
  11. -------------------
  12. Sequential floating forward selection for 5 features
  13. Accuracy using all features: 94.118%
  14. Accuracy using features ([1, 2, 12]): 94.118%
  15. -------------------
  16. -------------------
  17. Sequential forward selection
  18. Accuracy using all features: 82.353%
  19. Accuracy using features ([0, 1, 6, 10]): 88.235%
  20. -------------------
  21. -------------------
  22. Sequential backward selection
  23. Accuracy using all features: 64.706%
  24. Accuracy using features ([3, 5, 9, 10, 11]): 82.353%
  25. -------------------
  26. -------------------
  27. Sequential forward selection for max 10 features
  28. Accuracy using all features: 82.353%
  29. Accuracy using features ([0, 6, 7, 9]): 100.000%
  30. -------------------
  31. -------------------
  32. Sequential backward selection for max 10 Features
  33. Accuracy using all features: 47.059%
  34. Accuracy using features ([6, 9, 11]): 88.235%
  35. -------------------
  36. -------------------
  37. Sequential floating forward selection
  38. Accuracy using all features: 41.176%
  39. Accuracy using features ([1, 5, 8, 9, 10]): 76.471%
  40. -------------------