[Case Study]Regularized Discriminant Analysis using Julia

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Regularized Discriminant Analysis

---

Regularized discriminant analysis is the same as quadratic discriminant analysis with an additional parameter.

The additional parameter, lambda, represents the amount of shrinkage of each class covariance matrix towards the overall covariance matrix that is used in LDA. A value of lambda = 1 will result in linear discriminant analysis whereas a value of lambda = 0 will result in quadratic discriminant analysis. However, in both cases it is not recommended to use this generalized method and instead use the specialized methods lda or qda.

1. julia> rda_mod = rda(fm, iris[train,:], lambda=0.5)
   
2. Formula: Species ~ :(+(Sepal_Length,Sepal_Width,Petal_Length,Petal_Width))
   
3. 
   
4. Response:
   
5. 
   
6. 3x3 DataFrame:
   
7.                Group    Prior Count
   
8. [1,]        "setosa" 0.333333    37
   
9. [2,]    "versicolor" 0.333333    37
   
10. [3,]     "virginica" 0.333333    38
    
11. 
    
12. 
    
13. Lambda: 0.5
    
14. Gamma: 0
    
15. 
    
16. Class means:
    
17. 3x5 DataFrame:
    
18.                Group Sepal_Length Sepal_Width Petal_Length Petal_Width
    
19. [1,]        "setosa"      5.01622     3.48649      1.46757    0.243243
    
20. [2,]    "versicolor"      5.87297     2.77568      4.23514     1.32973
    
21. [3,]     "virginica"      6.68158     2.98158      5.59474     2.03421
    
22. 
    
23. 
    
24. 
    
25. julia> rda_pred = predict(rda_mod,iris[test,:])
    
26. 38x1 PooledDataArray{UTF8String,Uint32,2}:
    
27. "setosa"     
    
28. ⋮         
    
29. "virginica"
    
30. 
    
31. julia> 100*sum(rda_pred .== y[test])/length(y[test])
    
32. 100.0

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏1 回帖

关键词：discriminant Discriminan Case study Analysis discrim additional represents overall towards whereas