Data mining(DM)


Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Data mining(DM)

  1. 1. Data Mining : The Discovery Technology for Knowledge Management Yike Guo Dept. of Computing Imperial College
  2. 2. Course Overview <ul><li>Goal </li></ul><ul><ul><li>Basic Concepts of Data Mining </li></ul></ul><ul><ul><li>Basic Data Mining Techniques </li></ul></ul><ul><ul><li>Data Mining procedure in Real World Applications </li></ul></ul><ul><ul><li>Future Research Trends on Data Mining </li></ul></ul><ul><li>Reference Books </li></ul><ul><li>Advances in Knowledge Discovery and Data Mining U.M Fayyad and G, Piatetsky-Shapiro AAAI/MIT Press. 1996 </li></ul><ul><li>Predictive Data Mining: A Practical Guide Sholom M.Weiss and Nitin Indurkhya Morgan Kaufmann Publishers, Inc. 1997 </li></ul><ul><li>Data Mining Techniques Wiley Computer Publishing, 1997 </li></ul>
  3. 3. What does the data say? Day Outlook Temperature Humidity Wind Play Tennis 1 Sunny Hot High Weak No 2 Sunny Hot High Strong No 3 Overcast Hot High Weak Yes 4 Rain Mild High Weak Yes 5 Rain Cool Normal Weak Yes 6 Rain Cool Normal Strong No 7 Overcast Cool Normal Strong Yes 8 Sunny Mild High Weak No 9 Sunny Cool Normal Weak Yes 10 Rain Mild Normal Weak Yes 11 Sunny Mild Normal Strong Yes 12 Overcast Mild High Strong Yes 13 Overcast Hot Normal Weak Yes 14 Rain Mild High Strong No
  4. 4. Turing Data into Knowledge
  5. 5. Data Mining Machine Learning Statistics Databases High Performance & Distributed Computing Data Mining Infrastructure Enabling Technology Decision Support Knowledge Discovery
  6. 6. Why Data Mining <ul><li>Limitation of traditional database querying: </li></ul><ul><ul><li>Most queries of interest to data owners are difficult to state in a query language </li></ul></ul><ul><ul><ul><li>“ find me all records indicating fraud”=> “ tell me the characteristics of fraud” (Summarisation) </li></ul></ul></ul><ul><ul><ul><li>“ find me who likely to buy product X” (classification problem) </li></ul></ul></ul><ul><ul><ul><li>“ find all records that are similar to records in table X” (clustering problem) </li></ul></ul></ul><ul><ul><li>Ability to support analysis and decision making using traditional (SQL) queries become infeasible ( query formulation problem ). </li></ul></ul>
  7. 7. Relational Database Revisited <ul><li>Terabyte databases, consisting of billions of records, are becoming common </li></ul><ul><li>Relational data model is the defacto standard </li></ul><ul><li>A relational database : set of relations </li></ul><ul><li>A relation : a set of homogenous tuples </li></ul><ul><li>Relations are created, updated and queried using SQL </li></ul><ul><li>Query = Keyword based search </li></ul><ul><ul><li>SELECT telephone_number </li></ul></ul><ul><ul><li>FROM telephone_book </li></ul></ul><ul><ul><li>WHERE last_name = “Smith” </li></ul></ul>
  8. 8. SQL : Relational Querying Language <ul><li>Provides a well-defined set of operations: scan, join, insert, delete, sort, aggregate, union, difference </li></ul><ul><li>Scan -- applies a predicate P to relation R </li></ul><ul><ul><ul><li>For each tuple tr from R </li></ul></ul></ul><ul><ul><ul><li>if P(tr) is true, tr is inserted in the output stream </li></ul></ul></ul><ul><li>Join -- composes two relations R and S </li></ul><ul><ul><li>For each tuple tr from R </li></ul></ul><ul><ul><li>For each tuple ts from S </li></ul></ul><ul><ul><li>if join attribute of tr equals to join attribute of ts </li></ul></ul><ul><ul><li>form output tuple by concatenating tr and ts </li></ul></ul>
  9. 9. The Query Formulation Problem <ul><li>It is not solvable via query optimisation </li></ul><ul><li>Has not received much attention in the database field or in traditional statistical approaches </li></ul><ul><li>These problems are of inductive features: learning from data rather than search from data </li></ul><ul><li>Natural solution is via train-by-example approach to construct inductive models as the answers </li></ul>Consider the query : What kinds of weather condition are suitable for playing tennis ?
  10. 10. Why Data Mining Now <ul><li>Data Explosion </li></ul><ul><ul><li>Business Data : organisations such as supermarket chains, credit card companies, investment banks, government agencies, etc. routinely generate daily volumes of 100MB of data </li></ul></ul><ul><ul><li>Scientific Data: Scientific and remote sensing instruments collect data at the rates of Gigabytes per day: far beyond human analysis abilities. </li></ul></ul><ul><li>Data Wasting </li></ul><ul><ul><li>O nly a small portion (5% - 10%) of the collected data is ever analysed </li></ul></ul><ul><ul><li>Data that may never be analysed continues to be collected, at great expense. </li></ul></ul><ul><li>We are drowning in data, but starving for knowledge! </li></ul>
  11. 11. What is Data Mining Data Mining: a non-trivial data analysis process for identifying valid, useful and understandable patterns from databases.
  12. 12. <ul><li>Data: set of facts F ( records in a database) </li></ul><ul><li>Pattern : An expression E in a language L describing data in a subset FE of F and E is simpler than the enumeration of al l the facts of FE. FE is also called a class and E is also called a model or knowledge . </li></ul><ul><li>Data Mining Process : data mining is a multi-step process involving multiple choices, iteration and evaluation. It is non-trivial since there is no closed-form solution. It always involve intensive search. </li></ul><ul><li>Validity : E is true (with high probability) for F </li></ul><ul><li>Useful : patterns are not trivial inductive properties of data </li></ul><ul><li>Understandable: p atterns should be understandable by data owners to aid in understanding the data/domain </li></ul>
  13. 13. How Data Mining Works Historical Data (Data Warehouse) Predictive Models Operational Data Business Action Decision Evaluation Feedback Data Mining System Decision Support System Knowledge Business Intelligence Data
  14. 14. Data Warehousing <ul><li>“ A data warehouse is a subject-oriented, integrated, time-variant , and nonvolatile collection of data in support of management’s decision-making process.” --- W. H. Inmon </li></ul><ul><li>A data warehouse is </li></ul><ul><ul><li>A decision support database that is maintained separately from the organization’s operational databases. </li></ul></ul><ul><ul><li>It integrates data from multiple heterogeneous sources to support the continuing need for structured and /or ad-hoc queries, analytical reporting, and decision support. </li></ul></ul>
  15. 15. Modeling Data Warehouses <ul><li>Modeling data warehouses: dimensions & measurements </li></ul><ul><ul><li>Star schema : A single object (fact table) in the middle connected to a number of objects (dimension tables) radically. </li></ul></ul><ul><ul><li>Snowflake schema : A refinement of star schema where the dimensional hierarchy is represented explicitly by normalizing the dimension tables. </li></ul></ul><ul><ul><li>Fact constellations : Multiple fact tables share dimension tables. </li></ul></ul><ul><li>Storage of selected summary tables: </li></ul><ul><ul><li>Independent summary table storing pre-aggregated data, e.g., total sales by product by year. </li></ul></ul><ul><ul><li>Encoding aggregated tuples in the same fact table and the same dimension tables. </li></ul></ul>
  16. 16. Example of Star Schema Many Time Attributes Time Dimension Table Many Store Attributes Store Dimension Table Sales Fact Table Time_Key Product_Key Store_Key Location_Key unit_sales dollar_sales Yen_sales Measures Many Product Attributes Product Dimension Table Many Location Attributes Location Dimension Table
  17. 17. Example of a Snowflake Schema Many Time Attributes Time Dimension Table Many Store Attributes Store Dimension Table Sales Fact Table Time_Key Product_Key Store_Key Location_Key unit_sales dollar_sales Yen_sales Measures Supplier_Key Product Dimension Table Location_Key Location Dimension Table Product_Key Location_Key Location_Key Country Region Supplier_Key
  19. 19. View of Warehouses and Hierarchies <ul><li>Importing data </li></ul><ul><li>Table Browsing </li></ul><ul><li>Dimension creation </li></ul><ul><li>Dimension browsing </li></ul><ul><li>Cube building </li></ul><ul><li>Cube browsing </li></ul>
  20. 20. Construction of Data Cubes sum 0-20K 20-40K 60K- sum Comp_Method … ... sum Database Amount Province Discipline 40-60K B.C. Prairies Ontario All Amount Comp_Method, B.C. Each dimension contains a hierarchy of values for one attribute A cube cell stores aggregate values, e.g., count, sum, max, etc. A “sum” cell stores dimension summation values. Sparse-cube technology and MOLAP/ROLAP integration. “ Chunk”-based multi-way aggregation and single-pass computation. All, All, All
  21. 21. OLAP: On-Line Analytical Processing <ul><li>A multidimensional, LOGICAL view of the data. </li></ul><ul><li>Interactive analysis of the data: drill, pivot, slice_dice, filter. </li></ul><ul><li>Summarization and aggregations at every dimension intersection. </li></ul><ul><li>Retrieval and display of data in 2-D or 3-D crosstabs, charts, and graphs, with easy pivoting of the axes. </li></ul><ul><li>Analytical modeling: deriving ratios, variance, etc. and involving measurements or numerical data across many dimensions. </li></ul><ul><li>Forecasting, trend analysis, and statistical analysis. </li></ul><ul><li>Requirement: Quick response to OLAP queries. </li></ul>
  22. 22. OLAP Architecture <ul><li>Logical architecture: </li></ul><ul><ul><li>OLAP view: multidimensional and logic presentation of the data in the data warehouse/mart to the business user. </li></ul></ul><ul><ul><li>Data store technology: The technology options of how and where the data is stored. </li></ul></ul><ul><li>Three services components: </li></ul><ul><ul><li>data store services </li></ul></ul><ul><ul><li>OLAP services, and </li></ul></ul><ul><ul><li>user presentation services. </li></ul></ul><ul><li>Two data store architectures: </li></ul><ul><ul><li>Multidimensional data store: (MOLAP). </li></ul></ul><ul><ul><li>Relational data store: Relational OLAP (ROLAP). </li></ul></ul>
  23. 23. Dimension Browsing <ul><li>Product <====== </li></ul><ul><li>Location ======> </li></ul>
  24. 24. Decision Support with Data Warehouse <ul><li>Ad Hoc Queries: Q: How many customers do we have in London? A: 32776 </li></ul>
  25. 25. <ul><li>Report and Spreadsheet </li></ul>
  26. 26. <ul><li>OLAP: Q:What are the sales figures for Y in the different regions: </li></ul>
  27. 27. <ul><li>Statistics: Q: Is there a relation between age and buy behaviour? A: Older clients buy more </li></ul>
  28. 28. <ul><li>Data Mining: Q: What factors influence buying behaviour ? </li></ul><ul><ul><li>A1: : Young men in sports cars buy 3 times as much audio equipment (clustering/regression): </li></ul></ul><ul><ul><li>A2: Older woman with dark hair more often buy rinse (classification) </li></ul></ul><ul><ul><li>A3: Buyers of cars are also the buyers of houses (asociation) </li></ul></ul>Wage Old Young Middle Y N N N Y Hair color Age B W L H
  29. 29. Example Data Mining Applications <ul><li>Commercial : </li></ul><ul><ul><li>Fraud detection : Identify Fraudulent transaction </li></ul></ul><ul><ul><li>Loan approval: Establish the credit worthiness of a customer requesting a loan </li></ul></ul><ul><ul><li>Investment analysis : Predict a portfolio's return on investment </li></ul></ul><ul><ul><li>Marketing and sales data analysis: Identify potential customers; establishing the effectiveness of a sales campaign </li></ul></ul><ul><li>Medical: </li></ul><ul><ul><li>Drug effect analysis : from patient records to learn drug effects </li></ul></ul><ul><ul><li>Disease causality analysis </li></ul></ul><ul><li>Political policy: </li></ul><ul><ul><li>Election policy : people’s voting patterns </li></ul></ul><ul><ul><li>Social policy: tax/benefit policy </li></ul></ul><ul><li>Manufacturing: </li></ul><ul><ul><li>Manufacturing process analysis : identify the causes of manufacturing problems </li></ul></ul><ul><ul><li>Experiment result analysis : Summarise experiment results and create predictive models </li></ul></ul>
  30. 30. <ul><li>Scientific data analysis: </li></ul><ul><li>cataloguing in surveys, basic processing needed before higher-level science analysis can occur, scientific discovery over large data sets. </li></ul>Theory Experiments Simulation Data Assimilation (Data Warehousing) Data Mining (Statistical Computing and Machine Learning) Numerical Computing (Iterative Equation Solving) Numerical Computing : simulating the real world systems based on the underlying theory Data Assimilation : comprehending, consolidating and warehousing the simulation/experiment data Data Mining : analysis the warehoused simulation/experiment data for knowledge discovery
  31. 31. Related Fields: <ul><li>Machine learning: Inductive reasoning </li></ul><ul><li>Statistics : Sampling, Statistical Inference, Error Estimation </li></ul><ul><li>Pattern recognition: Neural Networks, Clustering </li></ul><ul><li>Knowledge Acquisition, Statistical Expert Systems </li></ul><ul><li>Data Visualisation </li></ul><ul><li>Databases: OLAP, Parallel DBMS, Deductive Databases </li></ul><ul><li>Data Warehousing: collection, cleaning of transactional data for on-line retrial </li></ul>