计算字符型变量的信息熵——macro

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

有空写了一个宏，可以直接计算字符型变量的信息熵；
raw_data是原始数据集，destination是分类变量；
关于数值型变量的信息熵算法有空继续；

1. %macro entropy(row_data,destination);
   
2. 
   
3. proc contents data=&row_data(drop=&destination) noprint out=var_list;run;
   
4. 
   
5. data var_cha_list;
   
6. set var_list;
   
7. where type=2;
   
8. keep name;
   
9. run;
   
10. 
    
11. /*first we need to get the character variable which we can calculate the entropy*/
    
12. 
    
13. %if %sysfunc(exist(var_cha_list)) ne 0 %then %do;
    
14. proc datasets lib=work nolist;
    
15. delete entropy_variable;
    
16. quit;
    
17. 
    
18. data entropy_variable;
    
19. length variable $40 entropy 8;
    
20. stop;
    
21. run;
    
22. %end;
    
23. 
    
24. /*create the table entropy_variable to store the results*/
    
25. 
    
26.         %let dsid = %sysfunc(open(var_cha_list));
    
27.         %if &dsid gt 0 %then %do;
    
28.         %let nobs = %sysfunc(attrn(&dsid,nobs));
    
29.         %do i=1 %to &nobs;/*start resolve this question, get the answers*/
    
30.                 %let rc = %sysfunc(fetch(&dsid,&i));
    
31.                 %let varnume = %sysfunc(varnum(&dsid,name));
    
32.                 %let variable = %sysfunc(getvarc(&dsid,&varnume));
    
33. ods html close;
    
34. ods output
    
35. CrossTabFreqs=pro;
    
36. proc freq data=&row_data;
    
37. table &variable * &destination;
    
38. run;
    
39. ods output close;
    
40. ods html;
    
41. 
    
42. /*use the freq proc to summary the data we need*/
    
43. 
    
44. data pro_temp(keep=&variable percent);
    
45. set pro(keep=&variable percent &destination);
    
46. where &destination='' and &variable^='';
    
47. run;
    
48. 
    
49. data pro_temp_2(keep=&variable rowpercent);
    
50. set pro(keep=&variable rowpercent &destination);
    
51. where &destination^='' and &variable^='';
    
52. run;
    
53. 
    
54. proc sql;
    
55. create table final as
    
56. select * from pro_temp as a
    
57. inner join pro_temp_2 as b
    
58. on a.&variable.=b.&variable.;
    
59. quit;
    
60. 
    
61. /*organize the data so we can use it directly*/
    
62. 
    
63. data final;
    
64. set final;
    
65. retain entropy 0;
    
66. if rowpercent^=0 and rowpercent^=100 then
    
67. entropy=entropy+rowpercent/100*log2(rowpercent/100)*percent/100;
    
68. else entropy=entropy;
    
69. run;
    
70. 
    
71. /*calculate the entropy*/
    
72. 
    
73. data entropy_&variable(keep=variable entropy);
    
74. retain variable;
    
75. set final end=eof;
    
76. variable="&variable";
    
77. entropy=-entropy;
    
78. if eof=1;
    
79. run;
    
80. 
    
81. proc append base=entropy_variable data=entropy_&variable force;run; /*get all the answer together*/
    
82.         %end;
    
83.         %let dsid = %sysfunc(close(&dsid));
    
84.         %end;
    
85. 
    
86. /*Worte by Albert.feng , if you use this code please note.*/
    
87. %mend entropy;
    
88. 
    
89. %entropy(c,play_golf);

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3 回帖

关键词：字符型变量 Macro Mac CRO acr 信息

%entropy(c,play_golf);
是我自己的测试，sorry，不该放在上面的

楼主好人，向你学习啊。

恩，不错

谢谢楼主，学习一下