OLAP Data Warehouse by fjzhangweiyun

VIEWS: 4 PAGES: 39

									   Background of
OLAP, DW, Data Mining
      Introductory Concepts
• Related Terms
  – OLAP
  – Data warehouse
  – Data mart
  – OLAP cube, multidimensional cube
  – Star schema
  – Fact, dimension
  – Dimensional modeling
  – Data mining
      Introductory Concepts
•Relevance of Terms

          OLAP
                          Data Mining
     Data Warehouse
                          Star Schema

  Multidimensional Cube
       Introductory Concepts
• OLAP(On-Line Analytical Processing)
  – Complex query processing for decision making
  – Report generation for the summary/aggregate
    /statistics over source data
  – Advanced data analysis techniques
• OLTP(On-Line Transaction Processing)
  – Simple transaction processing for regular (daily)
    business operations
  – Manipulating databases maintained for the
    organization’s activities
  – Optimized for simple operations (select,insert ...)
  Introductory Concepts

본부 현업 부서                총장, 교육부, 국회

       Transaction 요청

                         통계 자료 요청
 정보 시스템                  보고서 요청




업무 데이타베이스
  OLTP DB
       Introductory Concepts
• Solution (1)

   본부 현업 부서                   총장, 교육부, 국회

             Transaction 요청

                               통계 자료 요청
    정보 시스템          통계/보고서
                               보고서 요청
                    생성 모듈




  업무 데이타베이스
      OLTP DB
       Introductory Concepts
• Solution (2)

   본부 현업 부서                   총장, 교육부, 국회
                                    통계 자료 요청
             Transaction 요청
                                    보고서 요청
                                임의의 통계
    정보 시스템                       생성 도구

                                    OLAP 시스템


  업무 데이타베이스
      OLTP DB
       Introductory Concepts
• Solution (2)

   본부 현업 부서                   총장, 교육부, 국회
                                      통계 자료 요청
             Transaction 요청
                                      보고서 요청

    정보 시스템                      OLAP 시스템




  업무 데이타베이스                   분석용/요약 자료
      OLTP DB                 Data Warehouse
      Introductory Concepts
• OLAP systems/tools
  – Present summary/aggregate data for a large
    OLTP databases in several ways to decision
    makers
  – Manipulate DW over several dimensions
    dynamically to find useful information
• Data warehouse
  – A collection of data which are used for OLAP
  – Summary/aggregate data extracted from source
    OLTP databases which can support every
    possible OLAP requests.
  – Represented by star schema
     Introductory Concepts
•Relevance of Terms

         OLAP
                         Data Mining

   Data Warehouse
                         Star Schema

 Multidimensional Cube
Page 11
       Introductory Concepts
• OLTP(Online Transaction Processing)
  – Typical daily business query and update processing
• OLAP (Online Analytical Processing)
  – Complex query processing or report generation
     • Advanced data analysis techniques
     • ROLAP: Relational OLAP
     • MOLAP: Multi-dimensional OLAP
• Data Warehouse
  – An enterprise-wide data repository for decision
    support
• Data Mart
  – A smaller targeted DW for a business process
• Star Schema
  – A DB structure for data warehouse
                Data Warehouse
• DW is an integrated repository of data that is put
  into a form that can be easily understood,
  interpreted, and analyzed by the people who need
  to use it to make decisions
• Data are extracted from operational systems, then
  cleansed, integrated, transformed, and aggregated,
  into a read-only database that is optimized for
  decision making
• Data Warehouse is a
   –   Subject-oriented
   –   Integrated
   –   Time-variant
   –   Non-volatile

   – Collection of data in support of management’s decision-
     making process (W.H. Inmon)
                Data Warehouse
• Motivation for DW
  – An enterprise-wide repository of data, information,
    knowledge, and meta-data
     • Gather all the information into a single place for in-depth
       analysis
     • Decouple such analysis from OLTP systems
  – Transform the data into information
     • Provide right information in the right format at the right time
  – Perform sophisticated analysis of data
     • Perform trend analysis, time series analysis, risk analysis, etc.
     • Perform DSS exploration such as alternative formation,
       alternatives testing, decision-making, etc.
     • Discover/visualize hidden facts, patterns, correlations, rules,
       exceptions using data mining techniques
          Data Warehouse
• OLTP queries
  – How many shoes did we sell last month?
  – What are the age, address, phone of a
    certain student (e.g., Hong Gildong)
             Data Warehouse
• OLAP queries
  – How many size 10 shoes in red did we sell last month in
    the Midwest, the Northeast, compared with the same
    month last year, actual vs. budget?
  – What are the top 25 brands, by products, styles, and
    regions, for this period for total US based on sales dollars?
  – How much promotional expenses did we spend on
    customers who purchased less than $100 worth of
    products?
  – How much discount should we offer to boost the sales
    volume significantly?
  – Find the correlation between buying patterns of products
    of type A and those of type B.
  – What are sales trends?
  – What percent of the market do we own?
  – How are our defect rates improving?
  – Are our profits are increasing or decreasing?
• Comparison of OLTP and Data Warehouse
       – DWs require a new query-centric view of the data
                  OLTP                                   Data Warehouses & OLAP

Purpose           Daily business support,                Decision support,
                  Transaction processing                 Analytic processing
User              Data entry clerk, administrator,       Decision maker, executives
                  developer
DB design         Application oriented                   Subject-oriented

DB design model   ER model                               Star, snowflake, multidimensional model

Data structures   Normalized, complex                    Denormalized, simple

Data              Current, up-to-date operational data   Historical
                  Atomic                                 Atomic and summarized
                  Isolated                               Integrated
Usage             Repetitive, Routine                    Ad-hoc

Update            Transaction constantly generate        Data is relatively static
                  New data, read/write                   Often refreshed weekly, read mostly
Response time     Subsecond to second                    Seconds, minutes, worse

Index types       B+ trees                               B+ trees, bitmap index, join index

Systems           Transaction throughput,                Query throughput,
requirements      Data consistency                       Data accurary
              Data Warehouse
• Example Benefits of Data Warehousing & Data Mining
  – Fast sophisticated report generations
  – 4-15 times faster delivery
  – Young men buy beer on Friday nights when they buy
    diapers
  – More athletic shoes are sold on Friday evenings and
    Saturdays than during the rest of the week combined
  – More athletic shoes are sold when white tube socks are
    prominently displayed as part of a 2-for-1 sale.
  – In a retail chain, potato chip purchases were accompanied
    by a soda purchase in half the cases. That figure increases
    to 75% when there is a marketing promotion.
  – Blue Cross found some providers had superior treatment
    success rates for some fatal diseases
  – Victoria Secret found a particular incentive was ineffective,
    saving $300K per week.
               Data Warehouse
• Database Design for DW
  – Objective of a DW
     • Creating a database optimized for decision support
  – Limitations of ER model for DW applications
     • Normalized data model support large numbers of
       transactions with very few records.
     • ER models tend to be very complex and difficult to navigate.
     • ER model identifies first entities, then relationships
  – Four basic requirements of a warehouse design
     •   The schema must be simple.
     •   The data must be clean, consistent, and accurate
     •   Query processing must be fast
     •   Load the data into the warehouse quickly
  – Two types of data representation in DW
     • Star Schema
     • Multi-dimensional array
                 Star Schema
• Dimensional Model (Star Schema)
  – A database schema for data warehousing
  – Initially developed to simplify SQL queries (by Ralph
    Kimball)
  – Consists of a few central fact tables and many dimen
    sion tables
     • Dimension = analysis criteria
     • Fact = measurements aggregated over dimensions
              to be analyzed
  – Simplifies end-user query processing and high query
    performance
  – Used to reduce joins by OLAP/Relational engines
  – Relatively few tables and well-defined join paths
Star Schema
                       Star Schema
• Dimensional Model (Star Schema) consists of:
   – Fact table:
      •   Stores all transactions or factual data that are analyzed
      •   Typically numeric measures
      •   From millions to more than billion rows
      •   Example: Revenue, Actuals, Budgets, Sales, Orders, Bookings, Claims
   – Dimension table:
      •   Attributes about facts
      •   Supports grouping, browsing, constraining
      •   Provides the entry points into the DW
      •   Example: Time, Customer, Promotions, Demographics, LifeStyles, Prod
          ucts, Stores, Markets
                   Star Schema
• Dimension hierarchy
  – Each dimension table has several attributes
  – These attributes may have a hierarchy
     • Different level of aggregation for fact data
  – User may want aggregate data by some level of the
    dimension hierarchy

                                               Year    State
                                               Month   County
                                                day     City

                                                       Street
                         Star Schema
• The Strengths of Dimensional Modeling (Kimball 98)
   – Provides a predictable, standard framework
      • Report writers, query tools, and UIs could take advantage of DM structure
      • Supports presentation and performance
      • Simplifies the understanding and navigation of metadata
   – Is robust against unexpected changes in user behavior
      • Logical design is independent of the use of the schema
      • All dimensions are equally entry points to the system
   – Can be changed gracefully
      •   Could add new unanticipated facts of the same grain
      •   Could add new dimensions
      •   Add new unanticipated dimensional attributes
      •   Break existing dimension records down to a lower level of granularity
   – Availability of common modeling situations
      •   Slowly changing dimensions
      •   Heterogeneous products
      •   Pay-in-advance databases
      •   Factless facts as in event-handling
   – Availability of administrative utilities and SW processes that manage
     and use aggregates
                    Star Schema
• Variations of Star Schema
  – Star Schema
     • All dimensions are denormalized
     • Wide dimensions and deep facts
  – Snowflake Schema
     • All dimensions are normalized into 3NF
              Star Schema
• Summary of DW and Star Schema
  – DW is databases specially maintained for
    analytical processing
  – DW is organized by facts and dimensions
  – Star schema is a way of representing DW schema


• Cube, Data Mining??
Page 27
         Multidimensional Cube
• Motivation of Cube
  – Similar OLAP queries can be executed repeatedly to find
    some valuable information.
  – Users want to view data in several different perspectives
  – It is time-consuming to calculate these data which use
    similar facts and dimensions
  – Cube is a special data structure to meet this kind of
    OLAP queries’ requirements
                                 dim1            dim7
  dim1           dim6            dim2    fact1
  dim2   fact1   dim7
  dim3   fact2   dim8            dim1    fact1   dim7
  dim4           dim9
         fact3
  dim5           dim10
                                 dim1    fact1
         Multidimensional Cube
• Multidimensional Cube
  – Maintain every possible pairs of dimensions and
    aggregation of fact data
  – Types of OLAP
     • ROLAP : maintaining cube by relational table
     • MOLAP : maintaining cube by multidimensional array
                                 dim2

  dim1           dim6                   dim7      Query1
  dim2   fact1   dim7
                                                  Query2
  dim3   fact2   dim8
                          dim1
  dim4           dim9
         fact3                                    Query3
  dim5           dim10
         Multidimensional Cube
• Multidimensional Cube
  – Maintain every possible pairs of dimensions and
    aggregation of fact data
  – Types of OLAP
     • ROLAP : maintaining cube by relational table
     • MOLAP : maintaining cube by multidimensional array


  dim1           dim6                             Query1
  dim2   fact1   dim7
                                                  Query2
  dim3   fact2   dim8
  dim4           dim9
         fact3                                    Query3
  dim5           dim10
      Multidimensional Cube
• Multidimension Cube
  – Consist of several dimensions and facts
  – Its schema can be represented by star schema
  – Two different perspectives
    • Cube is a part of data warehouses
    • Cube is a data structure for OLAP tool which is
      extracted from data warehouses
       Introductory Concepts
• OLAP Operations
  – Drill down. View the ata in a more specialized level within
    a dimension
     • Show me total sales for by quarter
     • Show me total sales for this quarter by department
     • Show me total sales for this quarter, dept 22, by state
  – Roll-up. Converse of drill-down
     • Drilling down is adding new headers from the dimension tables
     • Rolling up is moving row headers
  – Slice and dice. Slice (dice) data on columns (rows),
    eliminating the rest of the display. Keep a subset of
    dimensions for selected values.
  – Pivot. Switching the arrangement of the dimensions
  – Cross-tab. A matrix report with X and Y axis
  – Other Operations
     • Ranking and computed attribute definitions
     • Multiple granularity aggregates (e.g., cube operator)
      Introductory Concepts
•Relevance of Terms

         OLAP
                         Data Mining

    Data Warehouse
                         Star Schema

 Multidimensional Cube
          Data Warehouse
Page 34
          OLAP tools




Page 35
                Data Mining
• What is Data Mining?
  – Is the process of knowledge discovery (useful
    information, patterns, groups, etc.) from a large of
    data.
  – Is the principle of soring through large amounts of
    data and picking out relevant information (by
    Wikipedia).
  – Is the nontrivial extraction of implicit, previously
    unknown, and potentially useful information from
    data.
  – Is the science of extracting useful information from
    large data sets or databases.
  – Data may be stored in files, databases, webs, data
    warehouses, etc.
                    Data Mining
• Data Mining algorithm


 Large input data      Mining     Knowledge
                      Algorithm   Information




  Data              Multidimen-
                                  OLAP tools
Warehouse           sional Cube
                Summary
•Relevance of Terms

         OLAP
                          Data Mining

    Data Warehouse
                          Star Schema

 Multidimensional Cube
Page 39

								
To top