首页 > 代码库 > 《SAS编程与数据挖掘商业案例》学习笔记之十九

《SAS编程与数据挖掘商业案例》学习笔记之十九

继续《SAS编程与数据挖掘商业案例》学习笔记,本文侧重数据处理实践,包括:HASH对象、自定义format、以及功能强大的正则表达式

一:HASH对象

Hash对象又称散列表,是根据关键码值而直接进行访问的数据结构,是根据关键码值而直接进行访问的数据结构,

sas提供了两个类来处理哈希表,用于存储数据的hash和用于遍历的hiter,hash类提供了查找、添加、修改、删除等方法,hiter提供了用于定位和遍历的firstnext等方法。

优点:键值的查找是在内存中进行的,有利于提高性能;

              hash表可以在数据步运行时,动态的添加更新或删除观测;

              hash表中可以很快的定位数据,减少查找次数;

常用方法:

definekey:定义键

Definedata:定义值

definedone:定义完成,可以载入数据

add:添加键值,如在hash表中已存在,则忽略;

replace:如果健在hash表中存在,则替换,如果不存在则添加键值

remove:清除键值对

find:查找健值,如果存在则将值写入对应变量

check:查找键值,如果存在则返回rc=0,不修改当前变量的值;

output:将hash表输出到数据集

clear:清空hash表,但并不删除对象

equal:判断两个hash类是否相等

 

find方法的示例:

libname chapt12 ‘f:\data_model\book_data\chapt12‘;

data results;

 if _n_=0 then set chapt12.participants;                  

   if _n_ = 1 then do;

    declare hash h(dataset:‘chapt12.participants‘);    

    h.definekey(‘name‘);

    h.definedata(‘gender‘, ‘treatment‘);

    h.definedone();

  end;

   set chapt12.weight;

  if h.find() = 0 then

    output;

run;

 

hiter对象的引例:

data patients;

  length patient_id $ 16 discharge 8;

  input patient_id discharge:date9.;

datalines;

smith-4123 15mar2004

hagen-2834 23apr2004

smith-2437 15jan2004

flinn-2940 12feb2004

;

data _null_;

  if _n_=0 then set patients;

  declare hash ht(dataset:"patients",ordered:"ascending");

  ht.definekey("patient_id");

  ht.definedata("patient_id", "discharge");

  ht.definedone();

  declare hiter iter("ht");

  rc = iter.first();

  do while (rc=0);

    put patient_id discharge:date9.;

    rc = iter.next();

  end;

run;

declare hiter iter("ht");hashht定义了一个遍历器iter,之后调用first方法将遍历器定位到hash表的第一条观测,然后使用next方法遍历hash表中的所有记录并输出。

 

商业实战-两个数据集的合并:

    data both1(drop=rc);    

      declare hash plan ();   

   rc = plan.definekey (‘plan_id‘); 

   rc = plan.definedata (‘plan_desc‘); 

   rc = plan.definedone ();  

   do until (eof1) ;     

     set chapt12.plans end = eof1;

     rc = plan.add ();    

  end;

  do until (eof2) ; 

     set chapt12.members end = eof2;

     call missing(plan_desc);

     rc = plan.find (); 

     output;   

  end;

  stop;

run;

上述程序可以简化为:

data both2;

   length plan_id $3 plan_desc $20;

   if _n_ = 1 then do;

         declare hash h(dataset:‘chapt12.plans‘);

         h.definekey(‘plan_id‘);

         h.definedata(‘plan_desc‘);

         h.definedone();

         call missing(plan_desc);

      end;

   set chapt12.members;

   rc=h.find();

run;

二:format

自定义format

Proc Format;

    Value $ Sex_Fmt

    ‘F‘=‘

    ‘M‘=‘

    Other = ‘未知‘;

    Value Age_Dur

    Low-10="10岁以下"            

    11-13="11-13"

    14-<15="14-15"

    15-High="15岁以上";

Run;

应用:

Data  test;

Set  sashelp.class(keep=sex age);

x=put(sex,$sex_fmt);y=put(age,age_dur.);

Run;

三:正则表达式:

/.../  一个正则表达式的起止;

|  数项之间的选择,“或”运算;

()   匹配组,标记一个子表达式的开始和结束位置;

   除换行符以外的任意字符;

\w  任一单词字符,数字大小写字母以及下划线

\W  任一非单词字符

\s   任一空白字符,包括空格、制表符、换行符、回车符、中文全角空格等;

\S   任一非空白字符,

\d   0-9任一数字

\D  任一非数字字符

[...]

[^...]

[a-z]  az

[^a-z]  不在从az范围内的任意字符

 匹配输入字符串的开始位置

 匹配输入字符串的结尾位置

\b  描述单词的前或后边界

\B  表示非单词边界

 匹配0次或多次

+ 匹配一次或多次

 匹配零次或 一次

{n}  匹配n

{n,}  匹配n次以上

{n,m}  匹配nm

 

常用函数:

Prxparse     定义一个正则表达式

Prxmatch  返回匹配模式的首次匹配位置

Call prxsubstr   返回匹配模式在目标字符串的开始位置和长度

Prxposn    返回正则表达式子表达式对应的匹配模式值

Call  prxposn    返回正则表达式子表达式对应的匹配模式和长度

Cal l  prxnext  返回匹配模式在目标字符串中的多个匹配位置和长度

Prxchange    替代匹配模式的值

Call prxchange   替代匹配模式的值

 

eg1

data _null_;

   if _n_ = 1 then pattern_num = rxparse("/cat/");

  

   retain pattern_num;

   input string $30.;

   position = rxmatch(pattern_num,string);

   file print;

   put pattern_num= string= position=;

datalines;

there is a cat in this line.

does not match cat

cat in the beginning

at the end, a cat

cat

;

run;

eg2:数据验证

data match_phone;

   set chapt12.phone_numbers;

   if _n_ = 1 then pattern = prxparse("/\(\d\d\d\) ?\d\d\d-\d{4}/");

   retain pattern;

   if prxmatch(pattern,phone) gt 0 then output;

run;

找出不匹配的手机号码

data unmatch_phone;

   set chapt12.phone_numbers;

   where not prxmatch("/\(\d\d\d\) ?\d\d\d-\d{4}/",phone);

run;

Eg3:提取匹配某种模式的字符串

data extract;

   if _n_ = 1 then do;

      pattern = prxparse("/\(\d\d\d\) ?\d\d\d-\d{4}/");

      if missing(pattern) then do;

         put "error in compiling regular expression";

         stop;

      end;

   end;

   retain pattern;

   length number $ 15;

   input string $char80.;

   call prxsubstr(pattern,string,start,length);

      if start gt 0 then do;

      number = substr (string,start,length); 

      number = compress(number," ");

      output;

   end;

   keep number;

datalines;

this line does not have any phone numbers on it

this line does: (123)345-4567 la di la di la

also valid (123) 999-9999

two numbers here (333)444-5555 and (800)123-4567

;

run;

eg4:提取名字

data ReversedNames;

   input name & $32.;

   datalines;

Jones, Fred

Kavich, Kate

Turley, Ron

Dulix, Yolanda

;

data FirstLastNames;

   length first last $ 16;

   keep first last;

   retain re;

   if _N_ = 1 then

      re = prxparse(‘/(\w+), (\w+)/‘);

   set ReversedNames;

   if prxmatch(re, name) then

      do;

         last = prxposn(re, 1, name);

         first = prxposn(re, 2, name);

      end;

run;

注:1,2分别代表正则表达式中的两个组

eg5:提取符合规定的名字

data old;

   input name $60.;

   datalines;

Judith S Reaveley

Ralph F. Morgan

Jess Ennis

Carol Echols

Kelly Hansen Huff

Judith

Nick

Jones

;

data new;

   length first middle last $ 40;

   re1 = prxparse(‘/(\S+)\s+([^\s]+\s+)?(\S+)/o‘);

   re2 = prxparse(‘/(\S+)(\s+)([^\s]+\s+)(?)(\S+)/o‘);

   set old;

   id1=prxmatch(re1, name);

   id2=prxmatch(re2, name);

   if id1 then

      do;

         first = prxposn(re1, 1, name);

         middle = prxposn(re1, 2, name);

         last = prxposn(re1, 3, name);

      end;

   if id2 then test=prxposn(re1, 4, name);

   put test=;

run;

Eg6:返回匹配模式的多个位置

data _null_;

   expressionid = prxparse(‘/[crb]at/‘);

   text = ‘the woods have a bat, cat, and a rat!‘;

   start = 1;

   stop = length(text);

   call prxnext(expressionid, start, stop, text, position, length);

      do while (position > 0);

         found = substr(text, position, length);

         put found= position= length=;

         call prxnext(expressionid, start, stop, text, position, length);

      end;

run;

注:首次执行call prxnext返回一个position,然后进入循环,在抽取满足条件的子串中,再次执行all prxnext,此时会返回下一个匹配的position

Eg7:替换文本

data cat_and_mouse;

   input text $char40.;

   length new_text $ 80;

   if _n_ = 1 then match = prxparse("s/[Cc]at/mouse/");

   retain match;

   call prxchange(match,-1,text,new_text,len,trunc,num);   

   if trunc then put "note: new_text was truncated";

datalines;

the Cat in the hat

there are two cat cats in this line

here is no replacement

;

run;

 

 

《SAS编程与数据挖掘商业案例》学习笔记之十九