DataOps, GitOps, and Docker containers are changing the role of Data Modeling, now at the center of end-to-end metadata management.
Success in the world of self-services analytics, data meshes, as well as micro-services and event-driven architectures can be challenged by the need to maintain interoperability of data catalogs/dictionaries with the constant evolution of schemas for databases and data exchanges.
In other words, the business side of human-readable metadata management must be up-to-date and in-sync with the technical side of machine-readable schemas. This process can only work at scale if it is automated.
It is hard enough for IT departments to keep in-sync schemas across the various technologies involved in data pipelines. For data to be useful, the business users must have an up-to-date view of the structures, complete with context and meaning.

In this session, we will review the options available to create the foundations for a data management framework providing architectural lineage and curation of metadata management.

Tired of building logical models that aren’t used? See how you can use JSON logical models, created with Hackolade, to code-generate semantic views that translate production data from systems into business language for use by data analysts and data scientists. Adding custom properties to Hackolade allows you to map between raw source system data copies in your EDW and desired business terms. Instead of huge, flat tables like “ACCOUNT” from Sales Cloud, expose “CUSTOMER” and even introduce logic not available in source data.

Objectives:

1. Enable auto-generated semantic views in the data warehouse to expose business terms.
2. Enable loose coupling between production systems and semantic views, absorbing at least some of the impact of production system changes.
3. Compare pass-through views to generating semantic views from logical JSON-based models.
4. Enable standards across logical models so that, for instance, the “ID” in the semantic view represents the logical key, regardless of the name in the underlying system.
5. Compare JSON attribute name grouping to flat naming or hand-coded views.

Rob Garrison is an experienced data modeler and architect. He currently works at Compassion International where he has the privilege of not managing anyone but spending his time on technical work. His previous roles were at Nike, DISH, WebMD, and Fiserv. Rob has presented at DMZ, PASS, and many local conferences and user groups. He also was the technical editor for Steve Hoberman’s book “Data Modeling for MongoDB: Building Well-Designed and Supportable MongoDB Databases” and has had articles published in Simple-Talk, Teradata Developer Exchange, and Database Journal.

The primary goal of this presentation is to demonstrate how data quality checks, predictive modeling techniques, and a knowledge graph was used to develop an effective credit card fraud detection framework. This comprehensive approach offers enhanced accuracy, robustness, and explainability in detecting fraudulent transactions.

The key stages of the framework are:

1. Data Cleansing to ensure the training dataset was free from errors, inconsistencies, and missing values. A high-quality training dataset was ready to be fed into the predictive model by eliminating these issues.
2. The predictive model comprises a sophisticated ensemble approach combined with multi-fold cross-validation techniques. By leveraging multiple models and evaluating their performance across different cross-validation folds, the accuracy and reliability of the fraud detection model was significantly enhanced. This approach allowed identifying patterns and anomalies associated with fraudulent transactions.
3. A knowledge graph was used to promote explainability and empower users to validate instances of fraud detection. Using the knowledge graph makes it possible to uncover hidden patterns, identify root causes, and easily validate the legitimacy of flagged transactions.

Learning objectives:

1. Implement data cleansing techniques to pre-process datasets for predictive modeling purposes effectively.
2. Employ critical thinking and evaluation skills to identify and select the optimal predictive model and feature set that yields the highest F1 score, indicating robust performance and accurate predictions.
3. Design and implement a user-friendly interface utilizing a knowledge graph to facilitate efficient data exploration and navigation, enabling users to navigate the dataset seamlessly.

Rafid is a data modeler who entered the field at the young age of 22, holding an undergraduate degree in Biology and Mathematics from the University of Ottawa. He was inducted into the DMC Hall of Fame by the Data Modeling Institute in July 2020, making him the first Canadian and 10th person worldwide to receive this honor. Rafid possesses extensive experience in creating standardized financial data models and utilizing various modeling techniques to enhance data delivery mechanisms. He is well-versed in data analytics, having conducted in-depth analyses of Capital Markets, Retail Banking, and Insurance data using both relational and NoSQL models. As a speaker, Rafid shared his expertise at the 2021 Data Modeling Zone Europe conference, focusing on the reverse engineering of physical NoSQL data models into logical ones. Rafid and his team recently placed second in an annual AI-Hackathon, focusing on a credit card fraud detection problem. Alongside his professional pursuits, Rafid loves recording music and creating digital art, showcasing his creative mind and passion for innovation in data modeling.

The emergence of online shopping post pandemic, has accelerated the need for ecommerce companies to understand the online behavior of their customers.  Focusing on the best practices of data modeling for data practitioners, this speech will demonstrate how to focus on the right customer profile, behavior, and site experience data to influence website design, customer taxonomies, and functionality decisions to optimize the customer experience.  Through the use of general customer behavior data analysis to machine learning capabilities, this speech will guide the data practitioner through the framework of success in this business model.  Leverage a real world case study as the artifact of the speech, the audience will be able to obtain a real-world use case in how to apply these customer data modeling best practices in e-commerce.

You will learn:

1. How to position the best data modeling method towards e-commerce experiences and customer taxonomies
2. How to assess the appropriate data correlation methods to influence online customer behavior strategy
3. How to identify how to provide effective insights into e-commerce journey and customer experience strategic decisions

Dr. Kyle Allison is a 20 year veteran in digital analytics, marketing & business.  He has spent time working with some of the most major ecommerce sites and retail companies in the country to include Best Buy, Dick’s Sporting Goods and The Exchange. Currently as a digital business and analytics consultant in the Doctor of Digital Strategy firm, he helps organizations balance effective digital business strategy planning with analytics & expert onsultation.  He is proficient in SAS, SPSS, Google Analytics, and Adobe Analytics. He is also a professor of Business Analytics at University of Texas at Dallas and Midland University.

Building semantic knowledge graphs that both deliver immediate value and evolve as your business model changes overtime can seem overwhelming — almost impossible. The most valuable knowledge graphs are flexible and dynamic, adapting to new data types, consumer patterns, and business requirements — but where do we start? In this session, Brian Platz will break down a simple formula that can jumpstart your Knowledge Graph journey: Model, Map, Connect, and Expand. We start small: identify a limited domain and model it with a semantic ontology. Map existing (or new) data to your model. Connect consumption patterns, applications, and business users to the data. Expand and evolve your data models, data consumers, and data sources using automation. Test often. Iterate. Automate.

We’ll discuss how the Model, Map, Connect, Expand formula can scale to new domains of business, allowing enterprise knowledge graphs to expand to new data and environments and evolve its models and requirements. By nature, the formula is circular, modular and interconnected: new data consumption requirements drive the need for new models, and new models can open up new analytical or operational capabilities.

We’ll end with a few case studies, showing how organizations use this formula as an ongoing strategy to transform proprietary data silos into semantic data ecosystems and future-proof their data domains for interoperability.

Brian Platz currently serves as Co-founder and Co-CEO at Fluree, a high-tech data management startup focused on providing trust, governance, and interoperability as native characteristics to data management. Brian was a successful entrepreneur and executive throughout the early internet days and SaaS boom, having founded the popular A-list apart web development community, along with a host of successful SaaS companies. He also was a fellow founding member of the Web Standards Project, a group formed to push Microsoft and Netscape to adhere to common standards.

Previous to establishing Fluree, Brian was the co-founder and general manager of SilkRoad technology’s Content Solutions, where he was responsible for providing all of the technology and services to deliver state-of-the-art web presences, content solutions, and collaborative content systems that allow a business to share information across its value chain. Before joining SilkRoad, Brian was the vice president of product management for divine, inc. from 2001 to 2003 where he led the product strategy and direction for divine’s widely recognized and award winning content management and collaboration products.

There are specialized tools (e.g. TARQL and SPARQL Anything) for converting data in traditional tables to RDF triples. The basic idea is to understand the meaning of each column and each table, and to find the terms in the ontology to express that meaning.   This process results in a mapping from the table schema to the ontology. This mapping is executed to convert the data in the relational table to triples.  Each cell in the table results in a single triple.

The triples collectively comprise the knowledge graph. The data may be queried producing a table of exactly what you want. The query language is called, SPARQL. 

Michael Uschold has over thirty years’ experience in developing and transitioning semantic technology from academia to industry. He pioneered the field of ontology engineering, co-authoring the first paper and giving the first tutorial on the topic in 1995 in the UK. 

As a senior ontology consultant at Semantic Arts since 2010, Michael trains and guides clients to better understand and leverage semantic technology.  He has built commercial enterprise ontologies in finance, insurance, healthcare, commodities markets, consumer products, electrical device specifications, manufacturing, corporation registration, and data catalogs.  The ontologies are used to create knowledge graphs that drive production applications.  This experience provides the basis for his book:  Demystifying OWL for the Enterprise, published in 2018.

During 2008-2009, Uschold worked at Reinvent on a team that developed a semantic advertising platform that substantially increased revenue. As a research scientist at Boeing from 1997-2008 he defined, led, and participated in numerous projects applying semantic technology to enterprise challenges. He received his Ph.D. in AI from The University of Edinburgh in 1991 and an MSc. from Rutgers University in Computer Science in 1982.

In Ensemble Modeling (Data Vault, Anchor Modelling, Focal Point, etc), we store all context in separate constructs. This presentation will reveal the secrets of its importance and why breaking up in multiple constructs is a benefit, not an issue! We will discuss the rule of 7, the years  1718 & 1892, why context might seem redundant while it’s not, and why history from a business perspective does not match with technical history. In other words, a whole session on how we capture context in Agile modeling. 

Remco is Vice President of International Programs at Genesee Academy. Remco is working in Business Intelligence and Enterprise Data Warehousing (EDW) with a focus on modeling and architecture including Ensemble and Data Vault modeling. He works internationally in Europe and is based in the Netherlands.
He had worked for several consulting companies in the Netherlands as a consultant in business intelligence from reporting, ETL, and modeling before starting his own companies, Coarem and BI Academy and joining Genesee Academy. He has years of experience teaching and speaking on modelling, business intelligence and data warehousing topics.

Before attempting to initiate digital improvements, your organization will need to take a look at its data-driven decision-making process. Aiken and Cesino provide executive insight into what a focused implementation looks like when an organization needs to make strategic decisions. The presentation will show how organizations rapidly realize concrete top- and bottom-line improvements directly attributable to maturing data practices. We will address key investments in technology, capabilities, and processes to achieve a sustained competitive advantage.

You will learn:

• How to identify key practices in support of a data strategy.
• How to align data – making it more available to inform business strategy.
• Where to look for process improvements.

Peter Aiken, PhD is an acknowledged Data Management (DM) authority, an Associate Professor at Virginia Commonwealth University, President of DAMA International, and Associate Director of the MIT International Society of Chief Data Officers.  For more than 35 years, Peter has learned from working with hundreds of data management practices in 30 countries including some of the world’s most important. Among his 12 books are the first on CDOs (the case for data leadership), focusing on data monetization,  on modern strategic data thinking and objectively specifying what it means to be data literate.  International recognition has resulted in an intensive schedule of events worldwide (pre-Covid).  Peter also hosts the longest running data management webinar series hosted by our partners at Dataversity.  Starting before Google, before data was big, and before data science, Peter has founded several organizations that have helped more than 200 organizations leverage data–specific savings have been measured at more than $1.5B USD. His latest is Anything Awesome.

As CEO of Visible Systems Corporation, Mr. Cesino enables companies to drive their business strategy to execution – viewing their enterprise from a “design thinking” perspective when creating new products and services.Additionally, through the award of a Small Business Innovative Research (SBIR) grant, he has led international eCommerce activities – providing technical input to the development and application of Models and Assessment Tools. Mr. Cesino holds an MBA from Suffolk University’s School of Management and degrees in Computer Science and Information Technology from Boston University and Northeastern University, respectively.