巧用二进制

当我们根据筛选数据中的可用信息时，有时候需要对某几个变量进行组合，根据特定组合事件的发生概率进行筛选，例如有A,B,C....N种事件，事件发生错误的次数未知，规定错误事件发生概率低于0.1%为安全事件，如果我们要判断哪几种事件所组成的事件集合为安全事件，那我们就必须要用到排列组合了。

 

思路如下：

要从N种事件中找出符合条件的事件组合，就要找到N种事件的所有可能的组合。所有可能的组合即2的N次方。可能的组合用2的N次方个观测遍历，每一个观测计算每一种可能性，问题即解。

SAS里面有BINARY.函数，用来将NUMERIC转换成二进制。再通过二进制的0/1进行判别，问题即解。

 

贴上CODE：

  1 %let XXX=C:\Users\yant07\XXX;
  2 libname XXX "C:\Users\yant07\XXX";
  3 
  4 %macro tmf( 
  5              date=1|2|3
  6             ,data=DATA  
  7             ,average=20
  8             );
  9 
 10 data _null_;
 11     daten=count("&date","|")+1;
 12     call symputx("daten",daten,G);
 13     datec=compress("&date",,"dk");
 14     call symputx("datec",datec,G);
 15 run;
 16 
 17 %do d= 1 %to &daten;
 18   %let date&d=%scan(%bquote(&date),&d,"|");
 19 
 20 data a&&date&d.; 
 21     set tmf.&data;
 22     keep did doc_submitter_NTID_from_Design tmf_item_id date y;
 23     rename doc_submitter_NTID_from_Design=id
 24            tmf_item_id=file;
 25     if ^missing(m_rule) then y=1;
 26     else y=0;
 27     date=input(scan(compress(put(QC_Task_Completed_Date,DATETIME27.6)),1,":"),anydtdte.);
 28     if month(date)=&&date&d.;
 29     format date date9.;
 30     proc sort nodupkey;
 31     by did id file;
 32 run;
 33 proc sql noprint;
 34     create table aa&&date&d. as
 35         select *,count(y) as nT&d
 36         ,sum(y) as nF&d
 37         from a&&date&d.
 38         group by id,file;
 39     proc sort;by id file descending date ;
 40     proc sort nodupkey;by id file;
 41 run;
 42 quit;
 43 
 44 /*modified by yant07 - count the average total file per each ID*/
 45 %end;
 46 data Merged_file;
 47     merge aa:;
 48     by id file;
 49     if ^missing(nT1) and ^missing(nT2) and ^missing(nT3) 
 50     and (nT1+nT2+nT3)/3>=&average.
 51     ;
 52 run;
 53 
 54 proc sql noprint;
 55     select distinct id ,count(distinct id) into: id separated by ",",:nid from Merged_file;
 56 quit;
 57 
 58 proc sql noprint;
 59     create table file_id as
 60     select id,count(distinct file) as file_id from Merged_file
 61     group by id;
 62     proc sort;by descending file_id; 
 63 quit;
 64 
 65 %do d= 1 %to &daten;
 66  %let date&d=%scan(%bquote(&date),&d,"|");
 67 
 68     %do i= 1 %to &nid;
 69      %let id&i = %scan(%bquote(&id) ,&i, ",") ;
 70     data &&id&i.&&date&d.;
 71         set Merged_file;
 72         where id="&&id&i";
 73         keep file id nt&d nf&d;
 74         proc sort nodupkey;
 75         by file id nt&d nf&d;
 76     run;
 77     proc sql noprint;
 78         select count(file) into : countfile      from &&id&i.&&date&d.;
 79         select file into: nfile separated by "," from &&id&i.&&date&d.;
 80         select nt&d into: ntfile separated by ","  from &&id&i.&&date&d.;
 81         select nf&d into: nffile separated by ","  from &&id&i.&&date&d.;
 82     quit;
 83     data q_&&id&i.&&date&d.;
 84         length combine $88.;
 85         array nfile[&countfile] $10;
 86         array ncfile[&countfile] $10;
 87         array ntfile[&countfile] 8;
 88         array nffile[&countfile] 8;
 89         retain nf nt combine;
 90         id="&&id&i.";
 91         comb=2**&countfile;
 92         fmt="Binary"||cats("&countfile")||".";
 93         do i = 1 to comb-1;
 94             binary=reverse(putn(i,fmt));
 95             nf=0;nt=0;combine="";
 96             do j = 1 to &countfile;
 97                 nfile[j]=substr(binary,j,1);
 98                 nCfile[j]=scan("&nfile",j,",");
 99                 nTfile[j]=scan("&ntfile",j,",");
100                 nFfile[j]=scan("&nffile",j,",");
101                 nf=nf+nfile[j]*nffile[j];
102                 nt=nt+nfile[j]*ntfile[j];
103                 if compress(nfile[j])=1 then do;
104                     combine=compress(combine)||"|"||compress(nCfile[j]);
105                 end;
106             end;
107             count=count(binary,"1");
108             rate=nf/nt;
109             if rate <=0.005 
110             then output;
111         end;
112         keep id binary rate combine count;
113         rename count=count&d.
114                binary=binary&d.
115                rate=rate&d.
116                ;
117         proc sort;by id combine ;
118     run;
119   %end;
120 %end;
121 
122 %do i= 1 %to &nid;
123   %let id&i = %scan(%bquote(&id) ,&i, ",") ;
124 
125 data qq_&&id&i.;
126     merge 
127     %do d= 1 %to &daten;
128       %let date&d=%scan(%bquote(&date),&d,"|");
129        q_&&id&i.&&date&d.(in=a&d)
130     %end; 
131     ;
132     if a1 and a2 and a3;
133     by id combine;
134 run;
135 %end;
136 data qqq;
137     set qq:;
138     keep id combine rate:;
139     if ^missing(combine);
140     rename combine=combination;
141     proc sort;by id combination;
142 run;
143 
144 ods printer pdf file="&tmf\TMF_Combinations_&datec._average&average..pdf"  ;
145 proc print data=qqq;
146     by id;
147 run;
148 ods printer close;
149 %mend;
150 
151 %tmf ;
152 *proc datasets library=work memtype=data kill nolist;
153 quit;

 

by yant07

posted @ 2016-10-24 17:33  yant07  阅读(409)  评论(0)  编辑  收藏  举报