With a data model, you can graphically model relationships between different types of information in companies. The data models help all users to store data and use it for a wide range of applications.
Data modeling describes the process of creating a visual representation of an information system or its parts to show relationships between data points and structures. The goal is to visualize the data used and stored in systems, as well as their relationships, groupings and organizations, formats and attributes.
What is a data model?
The basis for the creation of data models are the company's requirements. These are defined in advance through feedback from business partners to ensure that they are integrated into the design of a new system or taken into account when revising an existing system.
It is possible to model data at different levels of abstraction:
- The process begins by gathering information about business requirements from stakeholders and end users.
- These business rules are then translated into data structures in order to create a concrete database design.
- The data model can be compared to a roadmap, an architect's design or any other formal diagram that provides a deeper understanding of what is being designed.
Data modeling uses standardized schemas and formal techniques to create a uniform, consistent and predictable method for defining and managing data resources within an organization or beyond.
In the best case, data models are dynamic documents that continuously adapt to changing business needs.
They are central to supporting business processes and driving IT architecture and tactics planning. Data models can be shared with suppliers, collaboration partners and/or industry collaborators.
3 different data models
Every design process for databases and information systems starts with an abstract level and then becomes more and more concrete and specific. There are three main categories of data models, which differ in their level of abstraction.
The process starts with a conceptual model, then moves to a logical model, and finally ends with a physical model. These different types of data models are explained in detail below.
Conceptual Data Models
Conceptual data models, also known as domain models, provide comprehensive insight into the content, organization, and business rules of a system.
These models are usually created during the capture of the initial project requirements. They typically contain entity classes - which represent the relevant objects for the company in the data model - as well as their properties and restrictions.
In addition, the relationships between the entities as well as relevant security and data integrity requirements are considered. The representation of these models is mostly kept simple to ensure clear and convincing communication.
Logical data models
Logical data models provide a detailed representation of the concepts and relationships in a given domain as abstract models.
They are created using formal notation systems for data modeling and contain data attributes such as data types and relationships between entities. In contrast to technical system requirements, no technical details are specified in logical data models.
Although this phase is often skipped in agile or DevOps practices, logical data models can be especially useful in procedural implementation environments or in data-oriented projects such as data warehouse design or reporting system development.
Physical Data Models
Physical data structures provide a detailed concept for storing information in a database.
Unlike abstract approaches, they represent a concrete plan that can be implemented as a relational database.
They take into account associative tables that map the links between the various elements, as well as the necessary primary and foreign keys to maintain these links. Physical data models can also contain special features of a database management system (DBMS) in order to optimize performance.
Data modeling process in 6 steps
Data modeling requires a thorough evaluation of data processing and storage down to the smallest detail.
There are several specifications that determine which symbols are used to represent the data, how the models are constructed, and how the business requirements are communicated.
Regardless of this approach, formalized workflows provide an iterative sequence of tasks. These workflows typically include:
1. recognize the elements
The data modeling process starts with the recognition of the objects, operations, or ideas that are present in the data set to be designed. Each element should be coherent and logically distinct from all others.
2. identify the key characteristics of each entity
Each entity type can be distinguished from the others by having one or more specific properties called attributes.
For example, an entity named "Customer" might have attributes such as first name, last name, phone number, and salutation, while an entity named "Address" might include street name and number, city, state, country, and zip code.
3. find the connections between the elements
In the first conceptual phase of a data model, the way in which the various elements are linked is determined.
In the above example, each customer is located at a specific location. If an element called "Orders" were added to the model, each order would be connected to a specific address and would also be billed there. Typically, these connections are captured using the Unified Modeling Language (UML).
4. assign all properties to the corresponding elements
This ensures that the model correctly represents the company's use of the data. There are numerous established patterns for formal data modeling.
Developers in the object-oriented field often use analysis or design patterns, while stakeholders from other business areas may use alternative patterns.
5. distribute keys and choose a normalization level
This helps you balance redundancy reduction with performance requirements.
Normalization is a method of structuring data models (and their associated databases) by assigning numeric identifiers, called keys, to sets of data to represent connections between them without duplicating the data.
For example, if a key is assigned to each customer, this key can be linked to both the address and the order history without having to repeat this information in the table with the customer names.
Normalization usually requires less memory in the database, but this can be at the expense of query performance.
6. completion and review of the data model
The creation of data models is a recurring process that should be continuously optimized to adapt to changing business requirements.
Data Modeling Model Types
As database management systems have evolved, data modeling has also grown in complexity to meet the ever-increasing data storage requirements of enterprises.
Various types of models have emerged, including the following:
Hierarchical data models
Hierarchical data models are structured representations of one-to-many relationships organized in a tree-like format.
Each record has a single root or parent element that points to one or more child tables.
Although newer database models are more efficient, the hierarchical approach is still used in Extensible Markup Language (XML) systems and geographic information systems (GIS).
Relational databases
The advantages of relational modeling are that no detailed understanding of the physical properties of the data store is required and the data is linked through the use of tables. This reduces the complexity of the database.
Often relational databases use an ordered query language (SQL) for data management. Such databases are excellent for maintaining data integrity and reducing overlap. They are often used in cash registers and in various forms of transaction processing.
Entity-Relationship (ER) Data Models
Entity-relationship (ER) data models use precise diagrams to represent the connections between elements within a database.
A variety of ER modeling tools are used by data architects to generate descriptive representations that illustrate database design objectives.
Object-oriented data models
The relevant "objects" of object-oriented data models represent abstractions of real entities.
They are organized in class structures and have associated attributes. Object-oriented databases can contain not only tables, but also support more complex data links.
In practice, object-oriented databases are rarely used. Relational database systems dominate the application systems in business and research. With the strong spread of object-oriented programming languages, however, they are becoming increasingly important.
Compared to relational database systems, the performance of transactions decreases very quickly with the increase of data to be processed.
Due to the low distribution of object-oriented databases, there are only a few compatible interfaces.
This approach finds application in multimedia and hypertext databases, among others.
Dimensional data models
Dimensional data models can improve the speed of analysis in a data warehouse.
Unlike relational and ER models, which aim at efficient storage, dimensional models increase redundancyto make it easier to find information for reports and queries.
This type of modeling is usually used in OLAP systems in use.
There are two common models for structuring data in multiple dimensions.
The first is the Star Schemewhich divides the data into facts and dimensions. Each fact is surrounded by its associated dimensions, creating a star-shaped pattern.
The second model is the Snowflake scheme, which is similar to the star schema but contains additional layers of associated dimensions. This makes the branching pattern more complex.
Advantages of Data Models
Data modeling makes it easier for programmers, data architects, business analysts, and others involved to identify and understand the connections between information in a database or data warehouse.
It also offers the following advantages:
- Reduction of software and database development errors
- Increase consistency of documentation and system design across the enterprise
- Improve the performance of applications and databases
- Facilitate data mapping across the enterprise
- Optimization of communication between developers and business intelligence teams
- Simplify and accelerate database design at the conceptual, logical, and physical levels.
Data modeling providers and tools
There are a variety of commercial and open source options for computer-aided software engineering (CASE) that are widely used. These include various tools for data modelling, diagramming and visualization.
Other examples of Data Models include:
- Konfuzio stands out with an extremely flexible data modeling approach, which makes the software an attractive solution for companies that work with different data sources and structures. This flexibility comes from the ability to adapt to different data types, formats and storage systems without the need for rigid, predefined schemas.
- Erwin Data Modeler is a data modeling software based on the Integration DEFinition for Information Modeling (IDEF1X) language. The program now also allows the use of other notation methods, such as a dimensional approach.
- Enterprise Architect is an effective tool for visual modeling and design of enterprise information systems, architectures, software applications and databases. It uses object-oriented languages and standards and enables precise and detailed representation of the various components.
- The software ER/Studio enables effective database design and is compatible with many popular database management systems. It supports both relational and dimensional data modeling.
- There are a variety of free data modeling tools available, including open source alternatives such as. Open ModelSphere.
Conclusion
Thanks to the flexible application options of data models, you can optimally integrate them into your company. Not only will you save costs and reduce the error rate, but you will also optimize the performance of your databases and documentation, which will have a long-term impact on your company's figures and performance statistics. If you would like to read more about this topic, you can find suitable articles here:
- NLP - What is Natural Language Processing?
- Text Mining Wiki - Definitions and examples of use
- Process Mining: The most important definitions and tools
- Cloud hosting: Which enterprise solution is right for you?
EN 
DE 
ES 
FR