1 / 64

Core of Business “Intelligence” technology

Core of Business “Intelligence” technology. Database warehouse, data mining and on-line analytical processing . Business Intelligence and Analytics for Decision Support. The diagram show the role played by data warehouse, data-mining

niesha
Download Presentation

Core of Business “Intelligence” technology

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Core of Business “Intelligence” technology Database warehouse, data mining and on-line analytical processing

  2. Business Intelligence and Analytics for Decision Support The diagram show the role played by data warehouse, data-mining and olap in the “overall” business “decision making” process Business intelligence and analytics requires a strong database foundation, a set of analytic tools, and an involved management team that can ask intelligent questions and analyze data. Laudon and Laudon Chapter 10

  3. The Data Warehouse “A data warehouse is a subject-oriented, integrated, time-variant, and nonvolatile collection of “all” an organisation’s data in support of management’s decision making process.” • Data warehouses developed because E.G.: • if you want to ask “How much does this customer owe?” then the sales database is probably the one to use. However if you want to ask “Was this ad campaign more successful than that one?”, you require data from more disparate sources Other sources e.g. production, marketing etc.

  4. Characteristics of a Data Warehouse • Subject oriented – (based around business processes; e.g. sale of products,… • Integrated – inconsistencies removed • Nonvolatile – stored in read-only format • Time variant – data is “static” and update periodically; • Summarized – in decision-usable format; monthly average. • Large volume – data sets are quite large; all the pertinent data of an organisation • Non normalized – often redundant: star flake schema (it has dimension tables and fact tables):

  5. Customer Cust Purchases Product Ref Product Code ProdRef Eff. Date Customer ID Activity Date Product Code Customer ID Status Date ProdRef End Date Product Name Unit Price Product Category Product Type Product Sub Type Cust Addr State Cust ZIP Code Customer Type Customer Status ... Product Name Sales Rep ID Qty Purchased Total Dollars Promotion Flag Cust Averages Outlet Reference Customer ID Cust Average Date Sales Rep Ref Store ID Sales Rep ID Store Name Store Location Distribution Channel Cust Avg. End Date Cust Avg. Rev. Cust Longevity Sales Person Name Store ID The Atomic Schema

  6. Customer Location Selling Responsibility Cust ZIP Code Purchases 1 Sales Rep ID City State/Province Country Sales Rep ID Product Code Cust ZIP Code Customer Type Week Ending Date Sales Rep Name Store ID Store Name Store Location Sales Channel Customer Type Customer Type Days of Activity Unit Price Total Quantity Total Dollars Returned Qty Returned Dollars Promotion Qty Cust Type Desc Product Product Code Date Information Product Name Prod. Category Product Type Prod Sub Type Week Ending Date Month Quarter Year For Example:

  7. Elements of the building of a Data warehousing infrastructure A data warehouse process model

  8. Meta Data • A key concept behind D.W. is Meta Data. • Meta data is data about the data (which has come from the data sources) and shows what data is contained in the DW, where it came from, and what changes have been made to it. • The metadata are essential ingredients in the transformation of raw data into knowledge. They are the “keys” that allow us to handle the raw data. • For example, a line in a sales database may contain: 1023 K596 111.21 • This is mostly meaningless until we consult the metadata (in the data directory) that tells us it was store number 1023, product K596 and sales of $111.21.

  9. Meta Data Answers Questions for Users of the Data Warehouse • How do I find the data I need? • What is the original source of the data? • How was this summarization created? • What queries are available to access the data? • How have business definitions and terms changed over time? • How do product lines vary across organizations? • What business assumptions have been made?

  10. Dependent Data marts • A data mart is a data store that is subsidiary to a data warehouse of integrated data. • The data mart is directed at a partition of data (subject area) that is created for the use of a dedicated group of users and is sometimes termed a “subject warehouse” • The data mart might be a set of denormalised, summarised or aggregated data that can be placed on the data warehouse database or more often placed on a separate physical store. • Data marts can be “dependent data marts” when the data is sourced from the data warehouse. • Independent data marts represent fragmented solutions to a range of business problems in the enterprise, however, such a concept should not be deployed as it doesn’t have the “data integration” concept that’s associated with data warehouses.

  11. Independent Data marts • However, such marts are not necessarly all bad. • Often a valid solution to a pressing business problem: • Extremely urgent user requirements • The absence of a budget for a full data warehouse • The decentralisation of business units

  12. Data Warehousing Architecture • Access Tools • The principal purpose of the data warehouse is to provide information for strategic decision making. • The main Decision tools used to achieve this objective are: • Data mining tools • On-line analytical processing tools • Decision support systems / Executive information system tools

  13. Data Warehousing Typology • THE D.W. can be at single location i.e. a central data warehouse • The collection of data is replicated around multiple locations. This means users have a local copy of the data warehouse. This can improve query run-times, and reduce communications overheads. Distributed Data warehouse (Note: The principles associated with distributed database equally apply to Distributed Data warehouses, however, the static nature of the data needs to be factored in to the design process ) .

  14. Data Warehouse Construction Tips • Accept that your first try will require revision • Examine the data: What formats and specific data are needed to support your application? • Clean up the data before using it in the warehouse • Build a prototype mini-data warehouse as a learning experience and revise strategies as necessary • Plan on more users than anticipated wanting to use the warehouse • Keep storage requirements constantly in mind

  15. Sample type question • Discuss how D.W. can play’s key role in strategic decision making.

  16. Data Mining • The process of extracting valid, previously unknown, comprehensible, and actionable information from large databases and using it to make crucial business decisions. • Involves the analysis of data and the use of software techniques for finding hidden and unexpected patterns and relationships in sets of data.

  17. Data Mining • Data mining tools uses ,e.g. AI techniques, to help: • predict future trends: , • Segment datasets • “Product” association • allowing businesses to make proactive, knowledge-driven decisions.

  18. Data mining: A.I. techniques. • The most commonly used techniques A.I. techniques in data mining are: • Decision trees: Tree-shaped structures that represent sets of decisions. These decisions generate rules for the classification of a dataset. • Nearest neighbour method: A technique that classifies each record in a dataset based on a combination of the classes of the k record(s) most similar to it in a historical dataset. Sometimes called the k-nearest neighbour technique; a clustering technique • Rule induction: The extraction of useful if-then rules from data based on statistical significance. • Artificial neural networks: Predictive models that learn through training and resemble biological neural networks in structure.

  19. How Data Mining Works • For example, say that you are the director of marketing for a insurance company and you'd like to acquire some new customers • You could just randomly go out and mail coupons to the general population. However you would not achieve the required result. • Alternatively As the marketing director you have access to a lot of information about all of your customers: their age, sex, income range and credit card insurance.

  20. How Data Mining Works • The goal in prospecting is to make some decisions about the information in the lower right hand quadrant based on the model that we build going from Customer General Information to Customer Proprietary Information.

  21. An Algorithm for Building Decision Trees Consider the following using decision trees. The following is decision tree algorithm: 1. Let T be the set of training instances.2. Choose an attribute that best differentiates the instances in T.3. Create a tree node whose value is the chosen attribute. -Create child links from this node where each link represents a unique value for the chosen attribute. -Use the child link values to further subdivide the instances into subclasses. 4. For each subclass created in step 3: -If the instances in the subclass satisfy predefined criteria or if the set of remaining attribute choices for this path is null, specify the classification for new instances following this decision path. -If the subclass does not satisfy the criteria and there is at least one attribute to further subdivide the path of the tree, let T be the current set of subclass instances and return to step 2.

  22. How Data Mining Works • For instance, a simple model for a • Insurance company might be: • Customers who earn between 50 K to 60 K have a life insurance policy. • This model could then be applied to the general population to target those for the life insurance promotion. • The tree can be more complex e.g. See figure opposite

  23. Data Mining Operations • Data mining operations include: • Predictive modelling: decision trees, regression analysis… • Database segmentation: clustering techniques • Link analysis: decision trees, association rules

  24. Predictive Modeling Simple decision tree example • Applications of predictive modelling include direct marketing and use techniques like decision trees. • uses observations to form a model of the important characteristics of some phenomenon: e.g. those traits associated with those who will buy property

  25. Database Segmentation • Aim is to partition a database into an unknown number of segments, or clusters, of similar records. • Uses clustering techniques in order to group data • Applications of database segmentation include credit card fraud….

  26. Database Segmentation using a Scatterplot

  27. Link Analysis • Aims to establish links between records, or sets of records, in a database; one such example would be association discovery…. • Applications include product affinity analysis. • Finds items that imply the presence of other items in the same event.

  28. Link Analysis - Associations Discovery • Affinities between items are represented by association discovery. • e.g. ‘When a customer rents property for more than 2 years and is more than 25 years old, in 40% of cases, the customer will buy a property. This association happens in 35% of all customers who rent properties’.

  29. Examples of Applications of Data Mining • Retail / Marketing • Predicting response to mailing campaigns • Market basket analysis • Banking: • Detecting patterns of fraudulent credit card use. • Insurance • Claims analysis • Medicine • Identifying successful medical therapies for different illnesses

  30. Data mining in conclusion • Two critical factors for success with data mining are: • a large, well-integrated data warehouse and • a well-defined understanding of the business process within which data mining is to be applied (e.g. customer prospecting (target marketing), retention, campaign management etc.).

  31. Sample types questions • Discuss, using suitable examples how data mining can contribute to companies making a proactive knowledge driven decisions which could help with formulation of a companies strategy.

  32. What is OLAP • OLAP stands for "On-Line Analytical Processing.“ • OLTP ("On-Line Transaction Processing") • OLAP describes a class of technologies that are designed for live ad hoc data access and analysis. • OLTP generally relies solely on relational databases, • OLAP has become synonymous with multidimensional views of business data supported by multidimensional databases • Relational databases were never intended to provide data synthesis, analysis and consolidation functionality.

  33. What is OLAP • OLTP databases are optimised for transaction updating however, OLAP applications are used by managers and analysts for a higher level aggregate view of the data, thus they are designed for analysis. • Many problems that people try to solve using relational databases e.g. summaries are handled much more efficiently by an OLAP server than by RDBMS

  34. Key OLAP Features Although OLAP applications are found in widely divergent functional areas, as illustrate in the table opposite. Moreover they all have the following key features: multi-dimensional views of data (MD databases via Star Schema) Support complex calculations Time intelligence

  35. Star Schema: basis of MD view A star schema for credit card purchases

  36. Multi-dimensional view as a cube: also represented a 4 column table • Example of three-dimensional query. • What is the total amount and number of purchases for vehicles in region 2 for December. Multidimensional cube for credit card purchases

  37. Why Multidimensional Data • Queries requiring only a single number to be retrieved need not use multidimensional databases. • If queries involved retrieving multiple numbers and aggregating them for large databases can become intolerable as relational databases can scan only a few hundred records per second. • However multidimensional databases can add up 10,000 or more numbers in rows and columns per second. • Thus for such queries multidimensional databases have an enormous performance advantage

  38. Multi-dimensional Operations Slice – A single dimension operation Dice – A multidimensional operation Roll-up – A higher level of generalization Drill-down – A greater level of detail Rotation – View data from a new perspective

  39. Simple Hierarchies: Roll up • With hierarchical dimensions the database knows not to combine members of the dimension that are at different levels of the hierarchy: referred to as roll-up • It allows the user to view queries at all or any different levels e.g.. At street level ,city level, state level and region level. (refer to the above star schema example ) • Such hierarchies facilitate drill down to successive levels of detail: State level, city level, street level

  40. Multiple hierarchies: roll up • Utilising multiple hierarchies e.g. product sales can roll up by region, by type , by brand name and so forth. Without this capability an extra dimension would have to be created for each. • Another use of multiple hierarchies is for geographical dimensions e.g.:

  41. Drill down to core database • Most organisations now utilise relational databases as standard for their data warehouses. • Often there is no need to replicate all the data in the relational database into a MD database for OLAP. • Summary level data can be kept in the MD database and detailed data in the relational database.

  42. Drilling to relational data • To get a single number from a MD database takes the same time as it does from a relational database. • Thus it would be futile to individual customers into a MD database. But for summarised data a MD database is superior. • Thus ideally you should be able to drill down through the MD database into the relational database. • Such an approach is useful as most of data volume will reside at the detailed level and will thus not hinder queries of the higher levels

  43. Support for complex calculations • Important computational features of OLAP servers inlcude: • Independently dimensioned variables (IDV) • Statistical calculations • Consolidation speed • Vector Arithmetic

  44. OLAP calculations : Variables • Variables are numeric measures (facts) such as Sales, Cost, price…; dimensions include region, customer type, product… : i.e. fact table and dimension tables • OLAP servers can treat variables as a special dimension. So one can select only the relevant dimensions for each variable (IDV) . See next slide • Must provide a range of powerful computational and statistical methods such as that required by sales forecasting: regression analysis , projection . Correlations… • They can also incorporate various rules for consolidation

  45. Star schema for property sales of DreamHome

  46. Vector Arithmetic • Data held in 2-D arrays [Matrix] can be more easily manipulated than data stored in a relational table. • Thus a 2-D plane for actual can be easily subtracted from a plane from budget to give a plane for variance. • Such arithmetic allows entire planes of the database to be combined quickly.

  47. Time Series Data Types • Users want to look at trends in all aspects of their business e.g. sales trends, market trends etc. • A series of numbers representing a particular variable over time is called a time series e.g.. 52 weekly sales numbers is a time series. • Utilising a time-series data type allows you to store an entire string of numbers representing daily, weekly or monthly data. • Thus an OLAP server that supports time-series data type allows one to store historical data without having to specify a separate dimension for time. • Unlike other dimensions time has special attributes and rules.

  48. Time-series data type • Time series always have a particular periodicity. • Time series data must include rules to convert one periodicity to another • In the absence of a time-series data type a new dimension must be declared and labelled explicitly. • A time-series data cell contains a great deal of information compared with a single cell or even a full record.

More Related