Data modeling


Published on

Published in: Technology
1 Like
  • Be the first to comment

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

No notes for slide

Data modeling

  1. 1. Abdoulaye M Yansane, Abdoulaye Mouke Yansane, Mouke YansaneData Modeling Techniques for SuccessfulEvolutionary/Agile Database DevelopmentHome Search Agile DBAs Developers Enterprise Architects Enterprise Administrators Best PracticesThe goals of this article are to overview fundamental data modeling skills that alldevelopers should have, skills that can be applied on both traditional projects thattake a serial approach to agile projects that take an evolutionary approach. Mypersonal philosophy is that every IT professional should have a basicunderstanding of data modeling. They don’t need to be experts at data modeling,but they should be prepared to be involved in the creation of such a model, beable to read an existing data model, understand when and when not to create adata model, and appreciate fundamental data design techniques. This article is abrief introduction to these skills. The primary audience for this article isapplication developers who need to gain an understanding of some of the criticalactivities performed by an Agile DBA. This understanding should lead to anappreciation of what Agile DBAs do and why they do them, and it should help tobridge the communication gap between these two roles.Table of Contents1.What is data modeling?o How are data models used in practice?o What about conceptual models?o Common data modeling notations2.How to model datao Identify entity typeso Identify attributeso Apply naming conventionso Identify relationshipso Apply data model patternso Assign keyso Normalize to reduce data redundancyo Denormalize to improve performance3.Evolutionary/agile data modeling4. How to become better at modeling data1. What is Data Modeling?Data modeling is the act of exploring data-oriented structures. Like other modeling artifacts data models can be usedfor a variety of purposes, from high-level conceptual models to physical data models. From the point of view of an
  2. 2. object-oriented developer data modeling is conceptually similar to class modeling. With data modeling you identifyentity types whereas with class modeling you identify classes. Data attributes are assigned to entity types just as youwould assign attributes and operations to classes. There are associations between entities, similar to the associationsbetween classes – relationships, inheritance, composition, and aggregation are all applicable concepts in datamodeling.Traditional data modeling is different from class modeling because it focuses solely on data – class models allow youto explore both the behavior and data aspects of your domain, with a data model you can only explore data issues.Because of this focus data modelers have a tendency to be much better at getting the data “right” than objectmodelers. However, some people will model database methods (stored procedures, stored functions, and triggers)when they are physical data modeling. It depends on the situation of course, but I personally think that this is a goodidea and promote the concept in my UML data modeling profile (more on this later).Although the focus of this article is data modeling, there are often alternatives to data-oriented artifacts (never forgetAgile Modeling’s Multiple Models principle). For example, when it comes to conceptual modeling ORMdiagrams aren’t your only option – In addition to LDMs it is quite common for people to create UML classdiagrams and even Class Responsibility Collaborator (CRC) cards instead. In fact, my experience is that CRCcards are superior to ORM diagrams because it is very easy to get project stakeholders actively involved in thecreation of the model. Instead of a traditional, analyst-led drawing session you can instead facilitate stakeholdersthrough the creation of CRC cards.1.1 How are Data Models Used in Practice?Although methodology issues are covered later, we need to discuss how data models can be used in practice to betterunderstand them. You are likely to see three basic styles of data model:• Conceptual data models. These models, sometimes called domain models, are typically used to exploredomain concepts with project stakeholders. On Agile teams high-level conceptual models are often createdas part of your initial requirements envisioning efforts as they are used to explore the high-level static businessstructures and concepts. On traditional teams conceptual data models are often created as the precursor toLDMs or as alternatives to LDMs.• Logical data models (LDMs). LDMs are used to explore the domain concepts, and their relationships, ofyour problem domain. This could be done for the scope of a single project or for your entire enterprise.LDMs depict the logical entity types, typically referred to simply as entity types, the data attributesdescribing those entities, and the relationships between the entities. LDMs are rarely used on Agile projectsalthough often are on traditional projects (where they rarely seem to add much value in practice).• Physical data models (PDMs). PDMs are used to design the internal schema of a database, depicting thedata tables, the data columns of those tables, and the relationships between the tables. PDMs often prove tobe useful on both Agile and traditional projects and as a result the focus of this article is on physicalmodeling.Although LDMs and PDMs sound very similar, and they in fact are, the level of detail that they model can besignificantly different. This is because the goals for each diagram is different – you can use an LDM to explore
  3. 3. domain concepts with your stakeholders and the PDM to define your database design. Figure 1 presents a simpleLDM and Figure 2 a simple PDM, both modeling the concept of customers and addresses as well as the relationshipbetween them. Both diagrams apply the Barker notation, summarized below. Notice how the PDM shows greaterdetail, including an associative table required to implement the association as well as the keys needed to maintain therelationships. More on these concepts later. PDMs should also reflect your organization’s database namingstandards, in this case an abbreviation of the entity name is appended to each column name and an abbreviation for“Number” was consistently introduced. A PDM should also indicate the data types for the columns, such as integerand char(5). Although Figure 2 does not show them, lookup tables (also called reference tables or descriptiontables) for how the address is used as well as for states and countries are implied by the attributesADDR_USAGE_CODE, STATE_CODE, and COUNTRY_CODE.Figure 1. A simple logical data model.Figure 2. A simple physical data model.An important observation about Figures 1 and 2 is that I’m not slavishly following Barker’s approach to namingrelationships. For example, between Customer and Address there really should be two names “Each CUSTOMERmay be located in one or more ADDRESSES” and “Each ADDRESS may be the site of one or moreCUSTOMERS”. Although these names explicitly define the relationship I personally think that they’re visual noisethat clutter the diagram. I prefer simple names such as “has” and then trust my readers to interpret the name in eachdirection. I’ll only add more information where it’s needed, in this case I think that it isn’t. However, a significantadvantage of describing the names the way that Barker suggests is that it’s a good test to see if you actuallyunderstand the relationship – if you can’t name it then you likely don’t understand it.
  4. 4. Data models can be used effectively at both the enterprise level and on projects. Enterprise architects will oftencreate one or more high-level LDMs that depict the data structures that support your enterprise, models typicallyreferred to as enterprise data models or enterprise information models. An enterprise data model is one of severalviews that your organization’s enterprise architects may choose to maintain and support – other views may exploreyour network/hardware infrastructure, your organization structure, your software infrastructure, and your businessprocesses (to name a few). Enterprise data models provide information that a project team can use both as a set ofconstraints as well as important insights into the structure of their system.Project teams will typically create LDMs as a primary analysis artifact when their implementation environment ispredominantly procedural in nature, for example they are using structured COBOL as an implementation language.LDMs are also a good choice when a project is data-oriented in nature, perhaps a data warehouse or reportingsystem is being developed (having said that, experience seems to show that usage-centered approaches appear towork even better). However LDMs are often a poor choice when a project team is using object-oriented orcomponent-based technologies because the developers would rather work with UML diagrams or when the project isnot data-oriented in nature. As Agile Modeling advises, apply the right artifact(s) for the job. Or, as yourgrandfather likely advised you, use the right tool for the job. Its important to note that traditional approaches toMaster Data Management (MDM) will often motivate the creation and maintenance of detailed LDMs, an effortthat is rarely justifiable in practice when you consider the total cost of ownership (TCO) when calculating the returnon investment (ROI) of those sorts of efforts.When a relational database is used for data storage project teams are best advised to create a PDMs to model itsinternal schema. My experience is that a PDM is often one of the critical design artifacts for business applicationdevelopment projects.2.2. What About Conceptual Models?Halpin (2001) points out that many data professionals prefer to create an Object-Role Model (ORM), an example isdepicted in Figure 3, instead of an LDM for a conceptual model. The advantage is that the notation is very simple,something your project stakeholders can quickly grasp, although the disadvantage is that the models become largevery quickly. ORMs enable you to first explore actual data examples instead of simply jumping to a potentiallyincorrect abstraction – for example Figure 3 examines the relationship between customers and addresses in detail.For more information about ORM, visit 3. A simple Object-Role Model.
  5. 5. My experience is that people will capture information in the best place that they know. As a result I typically discardORMs after I’m finished with them. I sometimes user ORMs to explore the domain with project stakeholders butlater replace them with a more traditional artifact such as an LDM, a class diagram, or even a PDM. As ageneralizing specialist, someone with one or more specialties who also strives to gain general skills and knowledge,this is an easy decision for me to make; I know that this information that I’ve just “discarded” will be captured inanother artifact – a model, the tests, or even the code – that I understand. A specialist who only understands a limitednumber of artifacts and therefore “hands-off” their work to other specialists doesn’t have this as an option. Not onlyare they tempted to keep the artifacts that they create but also to invest even more time to enhance the artifacts.Generalizing specialists are more likely than specialists to travel light.2.3. Common Data Modeling NotationsFigure 4 presents a summary of the syntax of four common data modeling notations: Information Engineering (IE),Barker, IDEF1X, and the Unified Modeling Language (UML). This diagram isn’t meant to be comprehensive,instead its goal is to provide a basic overview. Furthermore, for the sake of brevity I wasn’t able to depict the highly-detailed approach to relationship naming that Barker suggests. Although I provide a brief description of eachnotation in Table 1 I highly suggest David Hay’s paper A Comparison of Data Modeling Techniques as he goesinto greater detail than I do.Figure 4. Comparing the syntax of common data modeling notations.
  6. 6. Table 1. Discussing common data modeling notations.Notation CommentsIE The IE notation (Finkelstein 1989) is simple and easy to read, and is well suited for high-level logical andenterprise data modeling. The only drawback of this notation, arguably an advantage, is that it doesnot support the identification of attributes of an entity. The assumption is that the attributes will bemodeled with another diagram or simply described in the supporting documentation.Barker The Barker notation is one of the more popular ones, it is supported by Oracle’s toolset, and is wellsuited for all types of data models. It’s approach to subtyping can become clunky with hierarchiesthat go several levels deep.IDEF1X This notation is overly complex. It was originally intended for physical modeling but has beenmisapplied for logical modeling as well. Although popular within some U.S. government agencies,particularly the Department of Defense (DoD), this notation has been all but abandoned by everyoneelse. Avoid it if you can.UML This is not an official data modeling notation (yet). Although several suggestions for a data modelingprofile for the UML exist, none are complete and more importantly are not “official” UML yet.However, the Object Management Group (OMG) in December 2005 announced an RFP for data-oriented models.3. How to Model DataIt is critical for an application developer to have a grasp of the fundamentals of data modeling so they can not onlyread data models but also work effectively with Agile DBAs who are responsible for the data-oriented aspects ofyour project. Your goal reading this section is not to learn how to become a data modeler, instead it is simply to gainan appreciation of what is involved.The following tasks are performed in an iterative manner:• Identify entity types• Identify attributes• Apply naming conventions• Identify relationships• Apply data model patterns• Assign keys• Normalize to reduce data redundancy• Denormalize to improve performanceVery good practical books about data modeling include JoeCelko’s Data & Databases and Data Modeling for InformationProfessionals as they both focus on practical issues with datamodeling. The Data Modeling Handbook and Data Model Patterns
  7. 7. are both excellent resources once you’ve mastered thefundamentals. An Introduction to Database Systems is a goodacademic treatise for anyone wishing to become a dataspecialist.3.1 Identify Entity TypesAn entity type, also simply called entity (not exactly accurate terminology, but very common in practice), is similarconceptually to object-orientation’s concept of a class – an entity type represents a collection of similar objects. Anentity type could represent a collection of people, places, things, events, or concepts. Examples of entities in anorder entry system would include Customer, Address, Order, Item, and Tax. If you were class modeling you wouldexpect to discover classes with the exact same names. However, the difference between a class and an entity type isthat classes have both data and behavior whereas entity types just have data.Ideally an entity should be normal, the data modeling world’s version of cohesive. A normal entity depicts oneconcept, just like a cohesive class models one concept. For example, customer and order are clearly two differentconcepts; therefore it makes sense to model them as separate entities.3.2 Identify AttributesEach entity type will have one or more data attributes. For example, in Figure 1 you saw that the Customer entityhas attributes such as First Name and Surname and in Figure 2 that the TCUSTOMER table had corresponding datacolumns CUST_FIRST_NAME and CUST_SURNAME (a column is the implementation of a data attribute within arelational database).Attributes should also be cohesive from the point of view of your domain, something that is often a judgment call. –in Figure 1 we decided that we wanted to model the fact that people had both first and last names instead of just aname (e.g. “Scott” and “Ambler” vs. “Scott Ambler”) whereas we did not distinguish between the sections of anAmerican zip code (e.g. 90210-1234-5678). Getting the level of detail right can have a significant impact on yourdevelopment and maintenance efforts. Refactoring a single data column into several columns can be difficult,database refactoring is described in detail in Database Refactoring, although over-specifying an attribute (e.g.having three attributes for zip code when you only needed one) can result in overbuilding your system and hence youincur greater development and maintenance costs than you actually needed.3.3 Apply Data Naming ConventionsYour organization should have standards and guidelines applicable to data modeling, something you should be ableto obtain from your enterprise administrators (if they don’t exist you should lobby to have some put in place). These
  8. 8. guidelines should include naming conventions for both logical and physical modeling, the logical namingconventions should be focused on human readability whereas the physical naming conventions will reflect technicalconsiderations. You can clearly see that different naming conventions were applied in Figures 1 and 2.As you saw in Introduction to Agile Modeling, AM includes the Apply Modeling Standards practice. The basicidea is that developers should agree to and follow a common set of modeling standards on a software project. Justlike there is value in following common coding conventions, clean code that follows your chosen coding guidelinesis easier to understand and evolve than code that doesnt, there is similar value in following common modelingconventions.3.4 Identify RelationshipsIn the real world entities have relationships with other entities. For example, customers PLACE orders, customersLIVE AT addresses, and line items ARE PART OF orders. Place, live at, and are part of are all terms that definerelationships between entities. The relationships between entities are conceptually identical to the relationships(associations) between objects.Figure 5 depicts a partial LDM for an online ordering system. The first thing to notice is the various styles applied torelationship names and roles – different relationships require different approaches. For example the relationshipbetween Customer and Order has two names, places and is placed by, whereas the relationship between Customerand Address has one. In this example having a second name on the relationship, the idea being that you want tospecify how to read the relationship in each direction, is redundant – you’re better off to find a clear wording for asingle relationship name, decreasing the clutter on your diagram. Similarly you will often find that by specifying theroles that an entity plays in a relationship will often negate the need to give the relationship a name (although someCASE tools may inadvertently force you to do this). For example the role of billing address and the label billed to areclearly redundant, you really only need one. For example the role part of that Line Item has in its relationship withOrder is sufficiently obvious without a relationship name.Figure 5. A logical data model (Information Engineering notation).
  9. 9. You also need to identify the cardinality and optionality of a relationship (the UML combines the concepts ofoptionality and cardinality into the single concept of multiplicity). Cardinality represents the concept of “how many”whereas optionality represents the concept of “whether you must have something.” For example, it is not enough toknow that customers place orders. How many orders can a customer place? None, one, or several? Furthermore,relationships are two-way streets: not only do customers place orders, but orders are placed by customers. This leadsto questions like: how many customers can be enrolled in any given order and is it possible to have an order with nocustomer involved? Figure 5 shows that customers place one or more orders and that any given order is placed byone customer and one customer only. It also shows that a customer lives at one or more addresses and that any givenaddress has zero or more customers living at it.Although the UML distinguishes between different types of relationships – associations, inheritance, aggregation,composition, and dependency – data modelers often aren’t as concerned with this issue as much as object modelersare. Subtyping, one application of inheritance, is often found in data models, an example of which is the is arelationship between Item and it’s two “sub entities” Service and Product. Aggregation and composition are muchless common and typically must be implied from the data model, as you see with the part of role that Line Item takeswith Order. UML dependencies are typically a software construct and therefore wouldn’t appear on a data model,unless of course it was a very highly detailed physical model that showed how views, triggers, or stored proceduresdepended on other aspects of the database schema.3.5 Apply Data Model PatternsSome data modelers will apply common data model patterns, David Hay’s book Data Model Patterns is the bestreference on the subject, just as object-oriented developers will apply analysis patterns (Fowler 1997; Ambler 1997)and design patterns (Gamma et al. 1995). Data model patterns are conceptually closest to analysis patternsbecause they describe solutions to common domain issues. Hay’s book is a very good reference for anyone involvedin analysis-level modeling, even when you’re taking an object approach instead of a data approach because hispatterns model business structures from a wide variety of business domains.3.6 Assign KeysThere are two fundamental strategies for assigning keys to tables. First, you could assign a natural key which is oneor more existing data attributes that are unique to the business concept. The Customer table of Figure 6 there wastwo candidate keys, in this case CustomerNumber and SocialSecurityNumber. Second, you could introduce a newcolumn, called a surrogate key, which is a key that has no business meaning. An example of which is the AddressIDcolumn of the Address table in Figure 6. Addresses don’t have an “easy” natural key because you would need to useall of the columns of the Address table to form a key for itself (you might be able to get away with just thecombination of Street and ZipCode depending on your problem domain), therefore introducing a surrogate key is amuch better option in this case.Figure 6. Customer and Address revisited (UML notation).
  10. 10. Lets consider Figure 6 in more detail. Figure 6 presents an alternative design to that presented in Figure 2, adifferent naming convention was adopted and the model itself is more extensive. In Figure 6 the Customer table hasthe CustomerNumber column as its primary key and SocialSecurityNumber as an alternate key. This indicates thatthe preferred way to access customer information is through the value of a person’s customer number although yoursoftware can get at the same information if it has the person’s social security number. The CustomerHasAddresstable has a composite primary key, the combination of CustomerNumber and AddressID. A foreign key is one ormore attributes in an entity type that represents a key, either primary or secondary, in another entity type. Foreignkeys are used to maintain relationships between rows. For example, the relationships between rows in theCustomerHasAddress table and the Customer table is maintained by the CustomerNumber column within theCustomerHasAddress table. The interesting thing about the CustomerNumber column is the fact that it is part of theprimary key for CustomerHasAddress as well as the foreign key to the Customer table. Similarly, the AddressIDcolumn is part of the primary key of CustomerHasAddress as well as a foreign key to the Address table to maintainthe relationship with rows of Address.Although the "natural vs. surrogate" debate is one of the great religious issues within the data community, the fact isthat neither strategy is perfect and youll discover that in practice (as we see in Figure 6) sometimes it makes senseto use natural keys and sometimes it makes sense to use surrogate keys. In Choosing a Primary Key: Natural orSurrogate? I describe the relevant issues in detail.3.7 Normalize to Reduce Data Redundancy
  11. 11. Data normalization is a process in which data attributes within a data model are organized to increase the cohesion ofentity types. In other words, the goal of data normalization is to reduce and even eliminate data redundancy, animportant consideration for application developers because it is incredibly difficult to stores objects in a relationaldatabase that maintains the same information in several places. Table 2 summarizes the three most commonnormalization rules describing how to put entity types into a series of increasing levels of normalization. Higherlevels of data normalization (Date 2000) are beyond the scope of this book. With respect to terminology, a dataschema is considered to be at the level of normalization of its least normalized entity type. For example, if all ofyour entity types are at second normal form (2NF) or higher then we say that your data schema is at 2NF.Table 2. Data Normalization Rules.Level RuleFirst normal form(1NF)An entity type is in 1NF when it contains no repeating groups of data.Second normal form(2NF)An entity type is in 2NF when it is in 1NF and when all of its non-keyattributes are fully dependent on its primary key.Third normal form(3NF)An entity type is in 3NF when it is in 2NF and when all of its attributes aredirectly dependent on the primary key.Figure 7 depicts a database schema in ONF whereas Figure 8 depicts a normalized schema in 3NF. Read theIntroduction to Data Normalization essay for details.Why data normalization? The advantage of having a highly normalized data schema is that information is stored inone place and one place only, reducing the possibility of inconsistent data. Furthermore, highly-normalized dataschemas in general are closer conceptually to object-oriented schemas because the object-oriented goals ofpromoting high cohesion and loose coupling between classes results in similar solutions (at least from a data point ofview). This generally makes it easier to map your objects to your data schema. Unfortunately, normalizationusually comes at a performance cost. With the data schema of Figure 7 all the data for a single order is stored in onerow (assuming orders of up to nine order items), making it very easy to access. With the data schema of Figure 7you could quickly determine the total amount of an order by reading the single row from the Order0NF table. To doso with the data schema of Figure 8 you would need to read data from a row in the Order table, data from all therows from the OrderItem table for that order and data from the corresponding rows in the Item table for each orderitem. For this query, the data schema of Figure 7 very likely provides better performance.Figure 7. An Initial Data Schema for Order (UML Notation).
  12. 12. Figure 8. A normalized schema in 3NF (UML Notation).
  13. 13. In class modeling, there is a similar concept called Class Normalization although that is beyond the scope of thisarticle.
  14. 14. 3.8 Denormalize to Improve PerformanceNormalized data schemas, when put into production, often suffer from performance problems. This makes sense –the rules of data normalization focus on reducing data redundancy, not on improving performance of data access. Animportant part of data modeling is to denormalize portions of your data schema to improve database access times.For example, the data model of Figure 9 looks nothing like the normalized schema of Figure 8. To understand whythe differences between the schemas exist you must consider the performance needs of the application. The primarygoal of this system is to process new orders from online customers as quickly as possible. To do this customers needto be able to search for items and add them to their order quickly, remove items from their order if need be, then havetheir final order totaled and recorded quickly. The secondary goal of the system is to the process, ship, and bill theorders afterwards.Figure 9. A Denormalized Order Data Schema (UML notation).
  15. 15. To denormalize the data schema the following decisions were made:1.To support quick searching of item information the Item table was left alone.2.To support the addition and removal of order items to an order the concept of an OrderItem table was kept,albeit split in two to support outstanding orders and fulfilled orders. New order items can easily be insertedinto the OutstandingOrderItem table, or removed from it, as needed.3.To support order processing the Order and OrderItem tables were reworked into pairs to handle outstandingand fulfilled orders respectively. Basic order information is first stored in the OutstandingOrder andOutstandingOrderItem tables and then when the order has been shipped and paid for the data is thenremoved from those tables and copied into the FulfilledOrder and FulfilledOrderItem tables respectively.Data access time to the two tables for outstanding orders is reduced because only the active orders are being
  16. 16. stored there. On average an order may be outstanding for a couple of days, whereas for financial reportingreasons may be stored in the fulfilled order tables for several years until archived. There is a performancepenalty under this scheme because of the need to delete outstanding orders and then resave them as fulfilledorders, clearly something that would need to be processed as a transaction.4.The contact information for the person(s) the order is being shipped and billed to was also denormalized backinto the Order table, reducing the time it takes to write an order to the database because there is now onewrite instead of two or three. The retrieval and deletion times for that data would also be similarlyimproved.Note that if your initial, normalized data design meets the performance needs of your application then it is fine as is.Denormalization should be resorted to only when performance testing shows that you have a problem with yourobjects and subsequent profiling reveals that you need to improve database access time. As my grandfather said, if itain’t broke don’t fix it.5. Evolutionary/Agile Data ModelingEvolutionary data modeling is data modeling performed in an iterative and incremental manner. The articleEvolutionary Development explores evolutionary software development in greater detail. Agile data modeling isevolutionary data modeling done in a collaborative manner. The article Agile Data Modeling: From DomainModeling to Physical Modeling works through a case study which shows how to take an agile approach to datamodeling.Although you wouldn’t think it, data modeling can be one of the most challenging tasks that an Agile DBA can beinvolved with on an agile software development project. Your approach to data modeling will often be at the centerof any controversy between the agile software developers and the traditional data professionals within yourorganization. Agile software developers will lean towards an evolutionary approach where data modeling is just oneof many activities whereas traditional data professionals will often lean towards a big design up front (BDUF)approach where data models are the primary artifacts, if not THE artifacts. This problem results from a combinationof the cultural impedance mismatch, a misguided need to enforce the "one truth", and “normal” politicalmaneuvering within your organization. As a result Agile DBAs often find that navigating the political waters is animportant part of their data modeling efforts.6. How to Become Better At Modeling DataHow do you improve your data modeling skills? Practice, practice, practice. Whenever you get a chance you shouldwork closely with Agile DBAs, volunteer to model data with them, and ask them questions as the work progresses.Agile DBAs will be following the AM practice Model With Others so should welcome the assistance as well as thequestions – one of the best ways to really learn your craft is to have someone as “why are you doing it that way”.You should be able to learn physical data modeling skills from Agile DBAs, and often logical data modeling skills aswell.
  17. 17. Similarly you should take the opportunity to work with the enterprise architects within your organization. As yousaw in Agile Enterprise Architecture they should be taking an active role on your project, mentoring your projectteam in the enterprise architecture (if any), mentoring you in modeling and architectural skills, and aiding in yourteam’s modeling and development efforts. Once again, volunteer to work with them and ask questions when you aredoing so. Enterprise architects will be able to teach you conceptual and logical data modeling skills as well as instillan appreciation for enterprise issues.You also need to do some reading. Although this article is a good start it is only a brief introduction. The bestapproach is to simply ask the Agile DBAs that you work with what they think you should read.My final word of advice is that it is critical for application developers to understand and appreciate the fundamentalsof data modeling. This is a valuable skill to have and has been since the 1970s. It also provides a commonframework within which you can work with Agile DBAs, and may even prove to be the initial skill that enables youto make a career transition into becoming a full-fledged Agile DBA.7. References and Suggested Online Readings• Agile/Evolutionary Data Modeling• Agile Database Best Practices• Agile Master Data Management (MDM)• Agile Modeling Best Practices• Choosing a Primary Key: Natural or Surrogate?• Comparing the Various Approaches to Modeling in Software Development• Data & Databases• Data Model Patterns• Data Modeling for Information Professionals• The Data Modeling Handbook• Database Modeling Within an XP Methodology (Ronald Bradford)• Initial High-Level Architectural Envisioning• Initial High-Level Requirements Envisioning• Introduction to Data Normalization• Logical Data Modeling: What It Is and How To Do It by Alan Chmura and J. Mark Heumann• On Relational Theory• The "One Truth Above All Else" Anti-Pattern• Prioritized Requirements: An Agile Best Practice• Survey Results (Agile and Data Management)• When is Enough Modeling Enough?
  18. 18. Data Modeling Techniques, Rules, and Diagram ConventionsSection 4 of the On-line Course:Learning the Cadastral Data Content StandardTechnical SectionsSections 4 through 8 are the sections of the Cadastral Data Content Standard educational course whichpresent detailed technical concepts about data models, crosswalks, translations, and maintenance of theStandard.Section 4 describes the entity relationship diagram and the definitions and relationships used in the Cadastral DataContent Standard, clarifying the data modeling conventions used in the Standards logical model. Please note thatdata modeling is a precise and detailed discipline, often requiring a good bit of effort to gain a working knowledge. Ifyou are new to data modeling, keep in mind that the information presented here in Section 4 may require some extratime and patience to understand.Topics in Section 4:• Overview of the Model• Logical Models vs Physical Models• The Content Standard versus a Physical Standard• Links and References to Information on Data ModelingOverview of the Cadastral Data Content Standard ModelThe Cadastral Data Content Standard model is an illustration of the objects in the Standard. The model is known as alogical model, and is illustrated in an entity relationship diagram (or E-R diagram).The logical model describes the definitions or semantics of the cadastral information referred to in the Standard. Anentity relationship diagram is a shorthand method for showing the associations among various objects in themodel, and the relationships between the objects.The entity relationship diagram illustrates the models objects, such as the entities, attributes, and the associations(see *Note below).A logical data model is not an implementation model. Implementation requires modifying the logical data model tobest fit operating software. This process, called denormalization, is the process of combining entities into tables in adatabase that optimize the database operation.See the diagram conventions discussion for more information about the E-R diagram used in the Cadastral DataContent Standard.
  19. 19. (* Note: The term "association" is used throughout the Cadastral Data Content Standard to refer to descriptions ofhow data entities are related to each other. Some people may be more accustomed to using the term "relationship",and may wish to substitute that term for "association" while investigating the sections which describe the datamodel.)Logical Models vs Physical ModelsThe following is a description of the differences between logical models and physical models. Data modelingprofessionals often note that there are varying ways of dealing with such details as keys, relationships, andnormalization. Accordingly, the description below has been kept as general as possible.Logical models depict the true relationships of attributes as they are grouped into entities, relating attributes toattributes and entities to entities. Logical models are not concerned with implementation, storage mechanisms, andredundancy of data. Logical models are usually normalized. Normalized means that every attribute is independent,that is, not dependent on any other attribute.Physical models are concerned with the implementation of logical models, and are designed to account for datastorage, indexes, how to retrieve data, and how keys are concatenated. Physical models strive to optimize logicalmodels according to how data are going to be used, such as for reports, data entry, and analysis. Physical models takeinto account the software that will be used, as well as whether the database will be relational, hierarchical ornetwork.Entities do not have to be the same between the logical model and the physical model. That is, in order toaccomodate efficient use of data, a physical model may have a greater or fewer number of entities than a logicalmodel. The physical model assigns lengths to attribute fields. A physical model is usually de-normalized, that is,attributes may be assigned values and dependencies with other attributes to support using the data. For example, anattribute can be derived for one or more other attributes. The attribute is used daily for reporting purposes so thederived attribute is stored in the data base to avoid daily recalculation.The Content Standard versus a Physical StandardThe Cadastral Data Content Standard is just that, a content standard. The Standard defines the kinds of entities,attributes, range of values, and logical relationships which can go into a cadastral database. The Standard does notdefine the actual structure of a database, and deals with none of the field definitions or software coding componentsof a physical design.For example, the cadastral standard provides a unique nation-wide identification of principal meridians. The namesof the principal meridians have been standardized and are listed in the Standard document. In a physical format for acounty or state that uses one of the principal meridians, it does not make sense to repeat that value for every record inthe county or state. In this case the physical format does not include the principal meridian as defined in the Standardin the database. The value for the principal meridian can be generated and added upon data transfer or exchange.In another example, an organization may decide they want their physical database to combine bearings and distancesand their units of measure in the same file as the record boundary. This might be done to accommodate a
  20. 20. computational package, to increase the ease of review of values, or to enhance search performance. The CadastralData Content Standard does not provide for this kind of physical database design and use.The physical structure of cadastral databases will be dealt with by the Cadastral Data Transfer Profile, which iscurrently in development, and is described in Section 6.Links and References to Information on Data ModelingFor more information on understanding data models, begin with the web sites for:• Applied Information Science• There is a commercial data modeling product from agpw, inc., known as Data Master. Though we have notreviewed it and cannot endorse the product, you may find it to be worth investigating.Published information on modeling includes:Bruce, T.A., Designing Quality Databases with IDEF1X Information Models, Dorset House, 1992.Chen, P.P.S., "The Entity-Relationship Model -- toward a unified view of data". ACM Transactions on DatabaseSystems 1, 1, March 1976.Jackson, Michael A., System Development. Prentice Hall International, Englewood Cliffs, New Jersey, 1983.Federal Information Processing Standards Publication 184, the Standard for Integration Definition for InformationModeling, U.S. Department of Commerce, Technology Administration, National Institute of Standards andTechnology, December 1993.One of the best short summaries of Bachman and Chen data modeling methods which we have found is inMcDonnell Douglas ProKit WORKBENCH Application Manual, Chapter 8, Data Modeler. This is a proprietarysoftware documentation manual, so as far as we know it is not a book available for purchase. Contact McDonnellDouglas (1-800-225-7760) if you are interested. (Note: In April 2002 it was pointed out to us that this document isno longer easily available from McDonnel Douglas, and thus may be difficult to find.)Surprisingly, there is virtually no widely accessible published information on the Charles Bachman method of datamodeling. A search on the subject of Bachman data modeling brought up the following articles:• Bachman Information Systems Data Base Management No. 6, The Entity Relationship Approach to LogicalData Base Design. Q.E.D. Monograph Series (Wellesley: Q.E.D. Information Science, Inc. 1977).• C.W. Bachman "Data Structure Diagrams" Journal of ACM SIGBDP Vol 1 No 2 (March 1969) pages 4-10.• McFadden & Hoffer _Database Management_, 3e, Benjamin Cummings, 1991, ISBN 0-8053-6040-9 or Date_An Introduction to Database Systems_, Volume 1, 5e, Addison Wesley, 1990, ISBN 0-201-51381-1• Charles W. Bachman: The Role Data Model Approach to Data Structures. 1-18, published in S. M. Deen, P.Hammersley: Proceedings International Conference on Data Bases, University of Aberdeen, July 1980.Heyden & Son, 1980, ISBN 0-85501-495-4.
  21. 21. This ends Course Sectin 4. Use the links below to return to the top of this page, or to go on to Section 5, or any of theother Modules.
  22. 22. A Comparison of Data Modeling TechniquesDavid C. Hay[This is a revision of a paper by the same title written in 1995. In addition to stylistic updates, this paperreplaces all the object modeling techniques with the UML – a new technique that is intended to replace atleast all these.]Peter Chen first introduced entity/relationship modeling in 1976 [Chen 1977]. It was a brilliant idea thathas revolutionized the way we represent data. It was a first version only, however, and many people sincethen have tried to improve on it. A veritable plethora of data modeling techniques have been developed.Things became more complicated in the late 1980’s with the advent of a variation on this theme called"object modeling". The net effect of all this was that there were now even more ways to model the structureof data. This was mitigated somewhat in the mid-1990s, with the introduction of the UML, a modelingtechnique intended to replace at least all the "object modeling" ones. As will be seen in this article, it is notquite up to replacing other entity/relationship approaches, but it has had a dramatic effect on the objectmodeling world.This article is intended to present the most important of these and to provide a basis for comparing themwith each other.Regardless of the symbols used, data or object modeling is intended to do one thing: describe the thingsabout which an organization wishes to collect data, along with the relationships among them. For this reason,all of the commonly used systems of notation fundamentally are convertible one to another. The majordifferences among them are aesthetic, although some make distinctions that others do not, and some do nothave symbols to represent all situations.This is true for object modeling notations as well as entity/relationship notations.There are actually three levels of conventions to be defined in the data modeling arena: The first issyntactic, about the symbols to be used. These conventions are the primary focus of this article. The seconddefines the organization of model diagrams. Positional conventions dictate how entities are laid out. Thesewill be discussed at the end of the article. And finally, there are conventions about how the meaning of amodel may be conveyed. Semantic conventions describe standard ways for representing common businesssituations. These are not discussed here, but you can find more information about them in books by DavidHay [1996] and Martin Fowler [1997]
  23. 23. These three sets of conventions are, in principle, completely independent of each other. Given any of thesyntactic conventions described here, you can follow any of the available positional or semantic conventions.In practice, however, promoters of each syntactic convention typically also promote at least particularpositional conventions.In evaluating syntactic conventions, it is important to remember that data modeling has two audiences.The first is the user community, that uses the models and their descriptions to verify that the analysts in factunderstand their environment and their requirements. The second audience is the set of systems designers,who use the business rules implied by the models as the basis for their design of computer systems.Different techniques are better for one audience or the other. Models used by analysts must be clear andeasy to read. This often means that these models may describe less than the full extent of detail available.First and foremost, they must be accessible by a non-technical viewer. Models for designers, on the otherhand must be as complete and rigorous as possible, expressing as much as possible.The evaluation, then, will be based both on the technical completeness of each technique and on itsreadability.Technical completeness is in terms of the representation of:o Entities and attributeso Relationshipso Unique identifierso Sub-types and super-typeso Constraints between relationshipsA technique’s readability is characterized by its graphic treatment of relationship lines and entity boxes,as well as its adherence to the general principles of good graphic design. Among the most important of theprinciples of graphic design is that each symbol should have only one meaning, which applies where ever thatsymbol is used, and that each concept should be represented by only one symbol. Moreover, a diagram shouldnot be cluttered with more symbols than are absolutely necessary, and the graphics in a diagram should beintuitively expressive of the concepts involved.. [See Hay 98.]Each technique has strengths and weakness in the way it addresses each audience. As it happens, mostare oriented more toward designers than they are toward the user community. These produce models that arevery intricate and focus on making sure that all possible constraints are described. Alas, this is often at theexpense of readability.This document presents seven notation schemes. For comparison purposes, the same example model ispresented using each technique. Note that the UML is billed as an "object modeling" technique, rather than asa data (entity/relationship) modeling technique, but as you will see, its structures is fundamentally the same.This comparison is in terms of each technique’s symbols for describing entities (or "object classes", for the
  24. 24. UML), attributes, relationships (or object-oriented "associations"), unique identifiers, sub-types andconstraints between relationships. The following notations are presented here.At the end of the individual discussions is your author’s argument in favor of Mr. Barker’s approach foruse in requirements analysis, along with his argument in favor of UML to support design.RelationshipsMr. Chen’s notation is unique among the techniques shown here in that a relationship is shown as a two-dimensional symbol — a rhombus on the line between two or more entities.Note that this relationship symbol makes it possible to maintain a "many-to-many" relationship withoutnecessarily converting it into an associative or intersect entity. In effect, the relationship itself is playing therole of an associative entity. The relationship itself is permitted to have attributes. Note how "quantity","actual price", and "line number" are attributes of the relationship Order-line in Figure 1.
  25. 25. Note also that relationships do not have to be binary. As many entities as necessary may be linked to arelationship rhombus.Cardinality/optionalityIn Mr. Chen’s original work, only one number appeared at each end, showing the maximumcardinality. That is, a relationship might be "one to many", with a "1" at one end and a "n" at the other.This would not indicate whether or not an occurrence of an entity had to have at least one occurrenceof the other entity.In most cases, an occurrence of an entity that is related to one occurrence of another must be relatedto one, and an occurrence of an entity that is related to more than one may be related to none, so mostof the time the lower bounds can be assumed. The event/event category model, however, is unusual.Having just a "1" next to event showing that an event is related to one event category would not showthat it might be related to none. The "n" which shows that each event category is related to more thanone event would not show that it must be related to at least one.For this reason, the technique can be extended to use two numbers at each end to show the minimumand maximum cardinalities. For example, the relationship party-order between purchase order andparty, shows 1,1 at the purchase order end, showing that each purchase order must be with no lessthan one party and no more than one party. At the other end, "0,n" shows that a party may or may notbe involved with any purchase orders, and could be involved with several. The event/event categorymodel would have "0,1" at the event end, and "1,n" at the event category end.In an alternative notation, relationship names may be replaced with "E" if the existence of occurrencesof the second entity requires the existence of a related occurrence of the first entity.NamesBecause relationships are clearly considered objects in their own right, their names tend to be nouns.The relationship between purchase-order and person or organization, for example, is called order-line.Sometimes a relationship name is simply a concatenation of the two entity names. For example party-order relates party and purchase order.Entity and relationship names may be abbreviated.Unique identifiersA unique identifier is any combination of attributes and relationships that uniquely identify an occurrence ofan entity.While Mr. Chen recognizes the importance of attributes as entity unique identifiers [Chen 1977, 23], hisnotation makes no provision for showing this. If the unique identifier of an entity includes a relationship to asecond entity, he replaces the relationship name with "E", makes the line into the dependent entity an arrow,and draws a second box around this dependent entity. (Figure 2 shows how this would look if the relationship
  26. 26. to party were part of the unique identifier of purchas-order). This still does not identify any attributes that arepart of the identifier.Figure 2: Existence Dependent RelationshipSub-typesA sub-type is a subset of the occurrences of another entity, its super-type. That is, an occurrence of a sub-typeentity is also an occurrence of that entity’s super-type. An occurrence of the super-type is also an occurrenceof exactly one or another of the sub-types.Though not in Mr. Chen’s original work, this extension is described By Robert Brown [1993] and Mat Flavin[1981].In this extension, sub-types are represented by separate entity boxes, each removed from its super-type andconnected to it by an "isa" relationship. (Each occurrence of a sub-type "is a[n]" occurrence of the super-type.) The relationship lines are linked by a rhombus and each relationship to a sub-type has a bar drawnacross it. In Figure 1, for example, party is a super-type, with person and organization as its sub-types. Thusan order-line must be either a product or a service. This isn’t strictly correct, since an order line is the factthat a product or a service was ordered on a purchase-order. It is not the same thing as the product or servicethemselves.Constraints between relationshipsThe most common case of constraints between relationships is the "exclusive or", meaning that eachoccurrence of the base entity must (or may) be related to occurrences of one other entity, but not more thanone. These will be seen in most of the techniques which follow below.Mr. Chen does not deal with constraints directly at all. This must be done by defining an artificial entity andmaking the constrained entities into sub-types of that entity. This is shown in Figure 1 with the entitycatalogue item, with its mutually exclusive sub-types product and service. Each purchase order has an order-line relationship with one catalogue item, where each catalogue item must be either a product or a service.CommentsMr. Chen was first, so it is not surprising that his technique does not express all the nuances that have beenincluded in subsequent techniques. It does not annotate characteristics of attributes, and it does not show theidentification of entities without sacrificing the names of the relationships.
  27. 27. While it does permit showing multiple inheritance and multiple type hierarchies, the multi-box approach tosub-types takes up a lot of room on the drawing, limiting the number of other entities that can be placed on it.It also requires a great deal of space to give a separate symbol to each attribute and each relationship.Moreover, it does not clearly convey the fact that an occurrence of a sub-type is an occurrence of a super-type.
  28. 28. Live chat by BoldchatDM STAT-1 Consultings founder and President Bruce Ratner, Ph.D. has made the companythe ensample for Statistical Modeling & Analysis and Data Mining in Direct & DatabaseMarketing, Customer Relationship Management, Business Intelligence, and InformationTechnology. DM STAT-1 specializes in the full range of standard statistical techniques,and methods using hybrid statistics-machine learning algorithms, such as its patentedGenIQ Model© Data Mining, Modeling & Analysis Software, to achieve its Clients Goals- across industries of Banking, Insurance, Finance, Retail, Telecommunications, Healthcare,Pharmaceutical, Publication & Circulation, Mass & Direct Advertising, Catalog Marketing,Online Marketing, Web-mining, B2B, Human Capital Management, Risk Management, andNonprofit Fundraising. Bruce’s par excellence consulting expertise is clearly apparent as hewrote the best-selling book Statistical Modeling and Analysis for Database Marketing:Effective Techniques for Mining Big Data. (based on Amazon Sales Rank).Clients Goals include:• Results-Oriented: Increase Response Rates; Drive Costs Down and Revenue Up;Increase Customer Retention; Stem Attrition; Check Churn; Increase CustomerAffinity - Match Products with Customer Needs; Enhance Collections & RecoveryEfforts; Improve Risk Management; Strengthen Fraud Detection Systems; IncreaseNumber of Loans without Increasing Risk; Work Up Demographic- based MarketSegmentation for Effective Product Positioning; Perform Retail CustomerSegmentation for New Marketing Strategies; Construct New Business AcquisitionSegmentation to Increase Customer Base; Identify Best Customers: Descriptive,Predictive and Look-Alike Profiling to Harvest Customer Database; Increase Value ofCustomer Retention; Generate Business-to-Business Leads forIncrease Profitability; Target Sales Efforts to Improve Loyalty Among the MostProfitable Customers; Improve Customer Service by Giving Marketing and SalesBetter Information; Build CRM Models for Identifying High-value Responders; BuildCRM Models to Run Effective Marketing Campaigns; Improve Human Resource
  29. 29. Management - Retain the Best Employees; Optimize Price and Package Offerings;Right Offer at the Right Time with the Right Channel; Maintain Product Profitabilityand Support Effective Product Management; Increase the Yield of NonprofitFundraising Campaigns; Optimize Customer Loyalty; CRM for Cross-Sell and Up-Sellto Improve Response Rates and Increase Revenue; CRM Segmentation for TargetedMarketing; Workforce Optimization; Personalize Recommendations for Information,Products or Services; Credit Scoring to Control Risk; Retain Best Customers andMaximize Their Profits; Nonprofit Modeling: Remaining Competitive and Successful;Subprime Lender Short Term Loan Models for Credit Default and Exposure; RetailRevenue Optimization: Accounting for Profit-eating Markdowns; Nonprofit Modeling:Remaining Competitive and Successful; Detecting Fraudulent Insurance Claims;Demand Forecasting for Retail; Cross-Sell and Up-Sell to Improve Response Ratesand Increase Revenue; Credit Scoring for Controlling Risk; and so on.• Analytical Strategy: Build, Score and Validate Logistic Regression Models, OrdinaryRegression Models, Variant Regression-based Models, Decision-Tree Models,Machine-Learning Models, Quasi-Experimental Design Models, MarketingMix Optimization Models; Latent Class Models, Survival/Proportional HazardsModels, and Structural Equation Models, Machine-Learning Conjoint Analysis, and allother models in the data analysts tool kit for problem-solution approaches.o Model Types: Acquisition/Prospect Models, Retention Models, AttritionModels, LifetimeValue Models, Credit Risk Models, Response-ApprovalModels, Contact-Conversion Models, Contact-Profit Models, Customer-ValueBased Segmentation Models; Credit Scoring Models, Web-traffic Models,Balanced Scorecard Models, Cross-sell/Up-sell Models, Zipcode-basedModels, Blockgroup-based Models Decision-Tree Inventory Forecast Models,Models for Maximizing Profits from Solicitations, Mortgage and Credit CardDefault Models, Trigger Marketing Model, Fraud Detection: Beyond the Rules-Based Approach, Workforce Optimization Model, Collaborative FilteringSystems, and an assortment of results-related analytical strategies.• Analytical Tactics: Procedure When Statistical Model Performance is Poor; Proceduresfor Data that are Too Large to be Handled in the Memory of Your Computer;Procedures for Data that Are Too Large to be Handled in the Memory of YourComputer; Detecting Whether the Training and Hold-out Subsamples Represent theSame Universe to Insure that the Validation of a Model is Unbiased; Data Preparationfor Determining Sample Size; Data Preparation for Big Data; The Revised 80/20Rule for Data Preparation; Implement Data Cleaning Methods; Guide Proper Use ofthe Correlation Coefficient; Understand Importance of the Regression Coefficient;Effect Handling of Missing Data, and Data Transformations; High PerformanceComputing for Discovering Interesting and Previously Unknown Information in -credit bureau, demographic, census, public record, and behavioraldatabases; Deliverance of Incomplete and Discarded Cases; Make Use of OtherwiseDiscarded Data; Determine Important Predictors; Determine How Large a Sample is
  30. 30. Required; Automatic Coding of Dummy Variables; Invoke Sample Balancing;Establish Visualization Displays; Uncover and Include Linear Trends andSeasonality Components in Predictive Models; Modeling a Distribution with a Mass atZero; Upgrading Heritable Information; "Smart" Decile Analysis for IdentifyingExtreme Response Segments; A Method for Moderating Outliers, Instead ofDiscarding Them; Extracting Nonlinear Dependencies: An Easy, Automatic Method;The GenIQ Model: A Method that Lets the Data Specify the Model; Data MiningUsing Genetic Programming; Quantile Regression: Model-free Approach; MissingValue Analysis: A Machine-learning Approach; Gain of a Predictive InformationAdvantage: Data Mining via Evolution; and many more analytical strategy-relatedanalytical tactics.
  31. 31. The Banking Industry Problem-Solution: Reduce Costs,Increase Profits by Data Mining and ModelingBruce Ratner, Ph.D.</STRONGIn today’s slow-moving economy the banking industry is in tough competitive “boxing ring,” in which they aregetting hit with high customer attrition rates. And, achieving their goals – reduce costs and increase profit – is amatter of “survival of the fittest.” Fortuitously, their gargantuan volumes of transaction data gathered daily are thekey ingredient for achieving their goals. High performance computing for discovering interesting and previouslyunknown information within the gargantuan data is needed as part of a tactical analytical strategy to build models towin their goals. Traditional statistical approaches are virtually ineffectual at data mining, i.e., uncovering undetectedcost-reduction/profit-gaining predictive relationships. This knowledge is vitally necessary for building models forreducing costs, and increasing profits. The purpose of this article is to demonstrate the strength of the data miningmuscle of the genetic data-mining feature of the GenIQ Model©. I discuss case studies, which use the “body blows”of genetic data mining to produce victorious cost-reduction, and profit-gaining models. For an eye-opening previewof the 9-step modeling process of GenIQ, click here. For FAQs about GenIQ, click here.
  32. 32. The Banking Industry Problem-Solution: Reduce Costs,Increase Profits by Data Mining and ModelingBruce Ratner, Ph.D.</STRONGIn today’s slow-moving economy the banking industry is in tough competitive “boxing ring,” in which they aregetting hit with high customer attrition rates. And, achieving their goals – reduce costs and increase profit – is amatter of “survival of the fittest.” Fortuitously, their gargantuan volumes of transaction data gathered daily are thekey ingredient for achieving their goals. High performance computing for discovering interesting and previouslyunknown information within the gargantuan data is needed as part of a tactical analytical strategy to build models towin their goals. Traditional statistical approaches are virtually ineffectual at data mining, i.e., uncovering undetectedcost-reduction/profit-gaining predictive relationships. This knowledge is vitally necessary for building models forreducing costs, and increasing profits. The purpose of this article is to demonstrate the strength of the data miningmuscle of the genetic data-mining feature of the GenIQ Model©. I discuss case studies, which use the “body blows”of genetic data mining to produce victorious cost-reduction, and profit-gaining models. For an eye-opening previewof the 9-step modeling process of GenIQ, click here. For FAQs about GenIQ, click here.
  33. 33. Demand Forecasting for Retail:A Genetic ApproachBruce Ratner, Ph.D.Accurate demand forecasting is essential for retailers to minimize the risk of stores running out of a product, or nothaving enough of a popular brand, color or style. Preseason and in-season forecast errors account for 20 to 25percent of losses in sales. Traditional demand forecasting methods for all stock-keeping units (SKUs) across allstores and all geographies have an inherent weakness of no ability to data mine the volumes of time-series data at theSKU-level. The purpose of this article is to present a machine learning approach – the GenIQ Model© – for demandforecasting that has demonstrated superior results compared to the traditional techniques.For more information about this article, call Bruce Ratner at 516.791.3544,1 800 DM STAT-1, or e-mail at STAT-1 website visitors will receive my latest book Statistical Modeling and Analysis for DatabaseMarketing: Effective Techniques for Mining Big Data at a 33%-off discount plus shipping costs - just for theasking.