Smart Data Discovery Using SAS Viya

Powerful Techniques for Deeper Insights

Felix Liao

• Publisher: SAS Institute
• Year: 2020
• ISBN:
  9781635267242
• Topic: Informatik
• Subtopic: Data Mining

Gain Powerful Insights with SAS Viya!

Whether you are an executive, departmental decision maker, or analyst, the need to leverage data and analytical techniques in order make critical business decisions is now crucial to every part of an organization. Smart Data Discovery with SAS Viya: Powerful Techniques for Deeper Insights provides you with the necessary knowledge and skills to conduct a smart discovery process and empower you to ask more complex questions using your data. The book highlights key components of a smart data discovery process utilizing advanced machine learning techniques, powerful capabilities from SAS Viya, and finally brings it all together using real examples and applications.

With its step-by-step approach and integrated examples, the book provides a relevant and practical guide to insight discovery that goes beyond traditional charts and graphs. By showcasing the powerful visual modeling capabilities of SAS Viya, it also opens up the world of advanced analytics and machine learning techniques to a much broader set of audiences.

Chapter 1: Why Smart Data Discovery?

Introduction

Why Smart Data Discovery Now?

Who Is This Book For?

Chapter Overview

Introduction

As information workers, our ability to leverage data and extract insights in order to make critical business decisions is fundamental to our success as individuals and the organizations that we work for. Regardless of whether you are an executive, departmental decision maker, or an analyst, the need to leverage data and analytical techniques effectively in order make business decisions is now pervasive throughout every part of an organization.

From the organization’s perspective, the historical approach of relying solely on statisticians or decision support specialists to prepare and analyze data is no longer a workable approach. Organizations today must involve everyone in their analytics efforts – especially those closest to core business functions – to truly leverage data for maximum strategic and tactical advantage.

The good news for both us as individuals and the organizations that we work for is that tremendous advancements in computer hardware and software in recent years have allowed us to collect more data than ever before. Furthermore, with the addition of advanced analytics and machine learning capabilities, modern analytics tools are now easier to use and have never been more powerful. These shifts have made data more accessible and true self-service analytics a reality today.

One area of analytics that has made a significant impact in recent years is self-service data visualization. These easy-to-use data visualization and exploration tools enable any information workers today to assemble data rapidly, explore hypotheses visually, and find new insights quickly. Data visualization tools empower business users and accelerate the process of insight discovery by reducing the need for statisticians, data modelers, or IT specialists. By shifting the process of insight discovery closer to the business and subject matter experts, it has enabled more timely and relevant insights to be discovered and acted upon.

“The greatest value of a picture is when it forces us to notice what we never expected to see.”

– John Tukey

Not only have these new data visualization tools accelerated the process of insight discovery, they have also allowed business users to ask more complex and forward-looking questions. These powerful, visual-based data discovery tools have revolutionized the traditional business intelligence solution space and led the way in terms of self-service analytics. Never before has it been easier for individual users to explore and visualize data for powerful insight with such ease.

With growing awareness and understanding of advanced data visualizations techniques, business users are now increasingly asking more complex questions, conduct more forward-looking analysis and eager to move beyond basic charts and graphs for answers. Enter the era of smart data discovery and the rise of citizen data scientist. Smart data discovery extends beyond the realm of traditional charts and visualization techniques with embedded machine learning techniques and algorithms. This new, augmented approach to data discovery leverages new visualization frameworks and automated machine learning capabilities to empower a new generation of users often described as citizen data scientists. Smart data discovery enables deeper insight discovery and empowers these new citizen data scientists to conduct deeper investigation, ask forward-looking questions, and develop valuable predictive insights.

What Is a Citizen Data Scientist?

Gartner defines a citizen data scientist as “a person who creates or generates models that leverage predictive or prescriptive analytics, but whose primary job function is outside of the field of statistics and analytics.” According to Gartner and other industry analysts, citizen data scientists are “power users” who can perform both simple and moderately sophisticated analytical tasks that would previously have required more expertise. They provide a complementary role to expert data scientists.

While smart data discovery holds tremendous promise, a new approach is needed in order to fully release its potential. At the core of this new approach is the recognition that citizen data scientists will need to be equipped with a new set of knowledge and skills, including the following:

1. Business Domain Knowledge – Smart data discovery needs to be built on a deeper understanding of the relevant business context and problem domains. Smarter insights can only come from asking more relevant and intelligent questions.

2. Analytics Techniques – A high level of familiarity with various analytical techniques and principles is required. While a PhD in Statistics is not necessary, a sound understanding of fundamental statistics and machine learning principles and techniques will be needed.

3. Product Know-How – Finally, it is about having access to the right tools and the necessary skills to bring it all together in a timely manner.

As depicted in Figure 1.1 below, these three skill sets in isolation are valuable, but when combined to solve a specific business problem, a new level of analytics and insight can be achieved – in this case, the sum is greater than its parts.

Figure 1.1: Smart Data Discovery Requirements

This book will help you navigate these intersecting knowledge domains and empower you to ask more complex questions by illustrating the key components of a smart data discovery process. We will highlight fundamental statistical concepts and how to leverage the relevant features. Most importantly, we will also be using real examples and applications to bring these concepts together.

SAS Viya, the latest evolution of the SAS platform will be used to demonstrate these examples throughout this book. An introduction to relevant features and functionalities that are needed in a smart data discovery process will be provided, followed by an explanation on how to leverage and interpret the various charts and outputs from SAS Viya.

Smart data discovery has the potential to shift an organization’s overall analytic maturity, accelerate its analytical efforts, and create a much bigger analytics workforce. From an individual perspective, it has the potential to transform the way you view data, conduct data discovery processes, and think about how complex business problems can be solved. In many ways, we are just at the start of this revolution, and I am hopeful that this book will help you and your organization lead the way in terms of realizing the true potential of data and analytics.

Why Smart Data Discovery Now?

Traditional Business Intelligence (BI) and data visualization tools do a great job of slicing and dicing data in order to help answer questions such as what happened and what is happening. These tools can also provide valuable dashboards and reports for the purpose of insights sharing and communication. However, they cannot easily identify correlating factors or help predict future outcomes. These outcomes might include: which customers will respond to a promotion? Who will churn to a competitor? And when will a piece of equipment fail? In a modern environment where businesses are no longer content with simply analyzing data from the past, but instead would also like to gain insights into the future, a new approach is clearly needed.

Smart data discovery extends the traditional data visualization paradigm by marrying intuitive visualizations with predictive analytics techniques and machine learning algorithms. As illustrated in Figure 1.2, the use of these more advanced techniques changes the nature of analysis from reactive to proactive. It changes the perspective from backward-looking to forward-looking, and it empowers the analytics professionals to dig deeper, investigate root causes, and predict future outcomes.

Figure 1.2: From Reactive to Forward-Looking and Proactive

What is not widely known is that data visualization has always been a fundamental tool for expert data scientists. Visualization techniques are often used by expert data scientists for correlation analysis, model feature selection, and testing different hypotheses quickly. Techniques range from the use of box plots and histograms for general exploration to the use of decision tree and scatter plots for feature selection. These exploration steps often act as the precursor to more complex predictive modeling techniques. Smart data discovery extends these analytical techniques to a much broader audience using an intuitive and visual-oriented approach that does not require complex programming or deep statistical knowledge.

From an organization’s perspective, smart data discovery has the potential to minimize the challenges associated with resourcing and staffing that most organizations face today. In an environment where there is still severe shortage of experienced, expert data scientists, smart data discovery not only has the potential to empower more users to leverage advanced machine learning techniques, it can also reduce the friction between the expert data scientist and the broader analytics professional community.

Data science is widely recognized as a highly collaborative process that require inputs from multiple teams and different personas. Expert data scientists are important and valuable, but they are only one part of the puzzle. Expert data scientists typically have a strong math and programming background and have a high-level understanding of the business process and functions. They normally do not ha...

Table of contents

• Contents
• Preface
• About This Book
• About The Author
• Acknowledgments
• Chapter 1: Why Smart Data Discovery?
• Chapter 2: The Role of The Citizen Data Scientist
• Chapter 3: SAS Visual Analytics Overview
• Chapter 4: Data Preparation
• Chapter 5: Beyond Basic Visualizations
• Chapter 6: Understand Relationships Using Correlation Analysis
• Chapter 7: Machine Learning and Visual Modeling
• Chapter 8: Predictive Modeling Using Decision Trees
• Chapter 9: Predictive Modeling Using Linear Regression
• Chapter 10: Bring It All Together