블로그 이미지
보미아빠

카테고리

보미아빠, 석이 (528)
밥벌이 (16)
싸이클 (1)
일상 (1)
Total
Today
Yesterday

달력

« » 2025.9
1 2 3 4 5 6
7 8 9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30

공지사항

최근에 올라온 글

https://www.microsoft.com/en-us/server-cloud/products/sql-server/
http://www.gartner.com/technology/reprints.do?id=1-237UHKQ&ct=141016&st=sb

 

장점은 왠지 저런 기관에서 중요하게 생각하는 기능을 열심히 만들어 넣는듯?한 분위기이고 , 단점은 심각하다. 어플라이언스가 없고(DW제외), 분류는 엔터프라이즈 급 디비라고 하지만 사용자는 엔터프라이즈 어플리케이션에 잘 안쓴다. 2012부터 가격은 돌아버렸고, 아주 정말 아우~~~~ 장난치나...HA/DR 구현이 쉽지 않단다....? 이건 먼소리지?

 

아래 표로는 좋아 보이지만 향 후 데이터베이스 시장은 엔터프라이즈 시장과 비 엔터프라이즈 시장으로 길을 나누어 갈 것이고 비 엔터프라이즈 시장은 mysql 이나 다른 무료 데이터베이스를 쓰면 충분하고 (all flash storage 쓰면 속도 아주 좋음 latch 는 설계 잘 하면 회피 할 수 있음) 캐시가 필요한 일반적인 곳은 윗단에 캐싱 레이어만 따로 구축하면 된다.

 

그나마, 데이터가 중요한 게임사는 가격 정책에 밀려 이미 다 떠나버렸고, mssql을 아직 게임사에 쓰면 개발사가 능력이 없는 것으로 보이고, 커뮤니티 같은 분야는 쓸 이유가 없고, 은행은 또 엔터프라이즈 느낌 안나는 mssql 잘 안쓴다 이말이군..

 

DB 배우는 사람은 그나마 미션 크리티컬한 곳에서 mssql 쓰는 곳에 취직을 해야 한다는 것인데, 이런곳에 고만 고만한 사람은 안 뽑는다. 과연 살아남을까? 이제 db를 처음 배운다면 잘 선택해야 할 듯 하다. 나는 끝까지 해서 메인프레임 관리자 처럼 되어봐야지 하하하하하하

 

 

Microsoft

Headquartered in Redmond, Washington, U.S., Microsoft markets its SQL Server DBMS for the operational DBMS market, as well as Microsoft Azure SQL Database (a database platform as a service) and Microsoft Azure Tables. Microsoft now has in-memory row-store technology for transactions in SQL Server 2014.

Strengths
  • Market vision: Microsoft's market-leading vision consists of in-memory computing (SQL Server 2014 now has full transaction in-memory support), NoSQL (with a new document-store DBMS), cloud offerings (both cloud-only and hybrid cloud), use of analytics in transactions (HTAP) and support of mobility. Its vision for in-memory computing and putting the "cloud first" is ahead of its competitors.
  • Strong execution: Microsoft SQL Server is an enterprisewide, mission-critical DBMS capable of competing with products from the other large DBMS vendors. Gartner's 2013 market share data shows Microsoft taking second place from IBM in terms of total DBMS revenue.
  • Performance and support: Reference customers were very positive, with the performance of SQL Server, documentation, support, ease of installation and operation all rated highly. Only 7% reported problems with the DBMS overall.
Cautions
  • Lack of an appliance: Microsoft still lacks an appliance for transactions (one comparable to its SQL Server Parallel Data Warehouse appliance), whereas its major competitors (IBM, Oracle and SAP) all offer one.
  • Market image: Although SQL Server is an enterprise-class DBMS, Microsoft continues to struggle to dispel a perception of weakness in this area. Inquiries from Gartner clients demonstrate a continuing perception that SQL Server is not used for mission-critical enterprisewide applications — a view that inhibits wider use of SQL Server as a primary, enterprise-class DBMS.
  • HA/DR and pricing issues: Reference customers again found the pricing model for SQL Server unacceptable (they gave it the lowest overall rating of any vendor in this Magic Quadrant) and blamed the price changes that came with SQL Server 2012. Microsoft also received one of the lowest overall scores for ease of implementing HA/DR.

 

 

 

 

 

Posted by 보미아빠
, |

연속된 값을 discretized 로 구간을 만들 수 있다.

아래와 같이 설정하면 되고, 연봉정보가 다 다를때, 나이대를 구분해 작업하고자 할 때 불연속 버전을 만들수 있다.

 

 

 

데이터 속성의 종류는 아래와 같고 그 특징을 이해해야 한다.

https://msdn.microsoft.com/ko-kr/library/ms174572.aspx

Posted by 보미아빠
, |

1. 선형회귀분석은 종속 변수와 독립 변수 간의 관계를 계산하고 이 관계를 예측에 사용

 

2. Y=aX+b

   를 구하는데, 모든 요소와 관련된 오류 수의 합계가 최소가 될 때까지 a와 b를 조정하여 회귀 수식을 구한다.

 

3. msft 의 선형회귀분석은 msft 의사 결정 트리의 변형이다. (트리의 분기가 없도록 조정한 것)

 

입력 연속된 숫자만 사용할 수 있다.

출력 연속된 숫자만 예측할 수 있다.

 

adventurework 에서 dimCustomer 의 평균연봉을 구하고자 한다면 다음과 같이 실습하면된다.

해당테이블에 나이가 없고 생일만 있으니 계산된 열을 이용해 추가한다.

 

dsv 에서 테이블에 계산된 컬럼을 추가한다. 일반적인 TSQL 이 먹는다.

 

 

선형 회귀분석 마이닝 구조와 모델을 만든다. 

 

 

 

모델링 플래그를 설정한다. (컬럼의 수동 추가 삭제 등의 경우 자동으로 설정되지 않는다.)

 

 

알고리즘 매개 변수를 확인한다.

선형 회귀분석의 경우 특별히 건드릴 변수가 없다.

 

 

 

마이닝 모델 뷰에서 Yearly Income 을 계산한 수식을 얻을 수 있다.

Yearly Income = 57,403.129+468.492*(Age-53.379)

 

 

만약 20살의 수입이 해당 모델로 학습한 경우 얼마가 되는지 계산해보면

 

 

마이닝 모델 예측에서는 더 쉽게 할 수 있다.

아래와 같이 설정하고 왼쪽 상단의 쿼리 버튼을 누르면 된다.

테이블을 넣어서 할 수도 있고, 단일 쿼리도 가능하고 dmx 쿼리를 바로 뽑아 낼 수도 있다.

 

 

 단일 쿼리를 구한것

 

 

단일 쿼리의 DMX 쿼리 확인  

 

 

 

쿼리분석기에서 데이터베이스에서 DMX 쿼리를 하면 똑같은 결과를 얻을 수 있다.

아래는 마이닝 파라메터를 호출한 경우이다.

 

 SELECT MINING_PARAMETERS
FROM $system.DMSCHEMA_MINING_MODELS
WHERE MODEL_NAME = 'LR_Customer_Income'

 

 


SELECT FLATTENED MODEL_NAME,
    (SELECT ATTRIBUTE_VALUE, VALUETYPE
     FROM NODE_DISTRIBUTION
     WHERE VALUETYPE =7)
AS t
FROM LR_Customer_Income.CONTENT

 

 

 

 SELECT FLATTENED NODE_DISTRIBUTION as t
FROM LR_Customer_Income.CONTENT

 

 

 

70% 샘플링을 해서 12939 건이 있고 연봉 평균값은 57403원이다.

계수는 (7번) 468 이고 평균 연령은 53세이다. 그래서 수식이

Yearly Income = 57,403.129+468.492*(Age-53.379) 이 나온것이다.

 

 

해석은 아래 링크를 보면 쉽게 할 수 있다.

https://msdn.microsoft.com/ko-kr/library/cc645906.aspx

이렇게 쉬운걸 ... 진작 공부해둘걸..... MSDN 만세~ !!!

 

Posted by 보미아빠
, |

최근에 달린 댓글

최근에 받은 트랙백

글 보관함