Data Analytics and AI

Jay Liebowitz

• Publisher: Auerbach Publications
• Year: 2020
• eBook ISBN: 9781000094671
• Topic: Computer Science
• Subtopic: Artificial Intelligence (AI) & Semantics
• Index: Computer Science

Analytics and artificial intelligence (AI), what are they good for? The bandwagon keeps answering, absolutely everything! Analytics and artificial intelligence have captured the attention of everyone from top executives to the person in the street. While these disciplines have a relatively long history, within the last ten or so years they have exploded into corporate business and public consciousness. Organizations have rushed to embrace data-driven decision making. Companies everywhere are turning out products boasting that "artificial intelligence is included." We are indeed living in exciting times. The question we need to ask is, do we really know how to get business value from these exciting tools?

Unfortunately, both the analytics and AI communities have not done a great job in collaborating and communicating with each other to build the necessary synergies. This book bridges the gap between these two critical fields. The book begins by explaining the commonalities and differences in the fields of data science, artificial intelligence, and autonomy by giving a historical perspective for each of these fields, followed by exploration of common technologies and current trends in each field. The book also readers introduces to applications of deep learning in industry with an overview of deep learning and its key architectures, as well as a survey and discussion of the main applications of deep learning. The book also presents case studies to illustrate applications of AI and analytics. These include a case study from the healthcare industry and an investigation of a digital transformation enabled by AI and analytics transforming a product-oriented company into one delivering solutions and services. The book concludes with a proposed AI-informed data analytics life cycle to be applied to unstructured data.

Chapter 1

Unraveling Data Science, Artificial Intelligence, and Autonomy

John Piorkowski

Contents

1.1 The Beginnings of Data Science

1.2 The Beginnings of Artificial Intelligence

1.3 The Beginnings of Autonomy

1.4 The Convergence of Data Availability and Computing

1.5 Machine Learning the Common Bond

1.5.1 Supervised Learning

1.5.2 Unsupervised Learning

1.5.3 Reinforcement Learning

1.6 Data Science Today

1.7 Artificial Intelligence Today

1.8 Autonomy Today

1.9 Summary

References

Often in discussions on data science, artificial intelligence (AI), and autonomy, the terms become conflated. Recently, we have experienced hype cycles in data science, artificial intelligence, and autonomy. Although these fields share common technologies and algorithms, their history has evolved independently, and they employ different frameworks and address different real-world applications. This chapter explains the commonalities and differences in the fields of data science, artificial intelligence, and autonomy. First, we will provide a historical perspective for each of these fields, followed by an exploration of common technologies and current trends in each field.

1.1 The Beginnings of Data Science

Data collection and analysis have been around long before the advent of the computer. A notable example is the work of Matthew Fontaine Maury, who was known as the “Scientist of the Seas.” Maury was a pioneer in the field of ocean navigation during the mid-1800s.^* He joined the Navy at the age of 19, but a stagecoach accident forced him to give up traveling the seas and take an assignment with the Navy at the Depot of Charts and Instruments. The Depot of Charts and Instruments would later become the US Naval Observatory. By studying meteorology, collecting data from ship’s logs, and creating charts, Maury revolutionized our understanding of oceanography and marine navigation. Figure 1.1 illustrates his 1851 Trade Wind Chart of the Atlantic Ocean, which assisted ship captains at the time with their cross-Atlantic journeys.

* https://blogs.loc.gov/maps/2018/07/scientist-of-the-seas-the-legacy-of-matthew-fontaine-maury/

Figure 1.1 “Trade wind chart of the Atlantic Ocean,” by Matthew Fontaine Maury, 1851. Geography and Map Division, Library of Congress.

A second example involves security analysis created by Benjamin Graham and David Dodd at the Columbia Business School in the 1920s. Security analysis involves analysis of financial data to inform investment decisions. These professors coauthored the classic text “Security Analysis” (1934), which describes the technique. Long-time successful investors such as Warren Buffet have been stewards of this technique.

A modern history of data science enabled by computing is often credited to a paper by John Tukey in 1962 titled “The Future of Data Analysis.” In this paper, he describes procedures for analyzing data, interpreting results, and planning for the gathering of data, as well as the statistics that apply to these procedures. Tukey’s prophecy of data analysis motivated a shift from theoretical statistics and advocated for applied statistics to become data analytics. Tukey’s paper has been reviewed more recently and still stands as a foundation for modern data science (Mallows, 2006).

In 1974, Pete Naur published the “Concise Survey of Computer Methods” and repeatedly used the term data science, defining it as the “the science of dealing with data, once they have been established, while the relation of the data to what they represent is delegated to other fields and sciences.” Even with Naur’s publication, many credit William Cleveland with coining the term “data science” with his publication in 2001. In his paper, he advocates for a substantial change to the field of statistics. To reinforce a significant change, he advocated for a new field called data science. He asserted that data science should include the following:

• Multidisciplinary investigations
• Models and methods for data
• Computational systems
• Pedagogy for education
• Evaluation of tools
• Theoretical foundations

Cleveland’s paper is cited as the seminal data paper; however, the field did not gain popularity until the explosion of internet connectivity, the low cost of data storage, and the “Big Data” era. The Big Data term refers to large and complex data that cannot be addressed with traditional relational database tool sets.^*

* https://en.m.wikibooks.org/wiki/Data_Science:_An_Introduction/A_History_of_Data_Science

1.2 The Beginnings of Artificial Intelligence

Many authors trace the beginnings of AI to the work of Aristotle and Euclid, who promoted the idea of human intelligence being mechanized. The genesis of artificial intelligence using computers rose out of a workshop at Dartmouth in 1956. This spawned the first wave of artificial intelligence that extended into the mid-1970s. The first wave was characterized by symbolic reasoning (Gunning, 2017). Symbolic reasoning relies on rule-based engines and expert systems that require engineers creating knowledge for these systems. The field experienced the first AI winter in the mid-1970s due to several circumstances,^* which included leading AI researchers identifying weaknesses in AI approaches and a summary report on the state of AI research published by the British government, the Lighthouse Report, in 1973. AI winters refer to increased skepticism within the field and reduced investment. Research funding diminished until the field of AI experienced another boom in the 1980s. The research of John Hopfield (1982) and David Rumelhart et al. (1986) described neural networks with back propagation, which renewed interest in AI research. These papers are credited with adding back propagation to neural networks. However, the original concept of back propagation was published by Paul Werbos a decade earlier (Werbos, 1974). A second trend fueling the boom was an effort to commercialize AI products.^† A primary focus of new products was expert systems built upon rule-based engines. This was short-lived, as expert systems never achieved full acceptance due to the narrowness of the problems they could solve. So, the field experienced another winter in the early 1990s.

* https://towardsdatascience.com/history-of-the-first-ai-winter-6f8c2186f80b

† https://towardsdatascience.com/history-of-the-second-ai-winter-406f18789d45

In the mid-1990s, the field of AI shifted from knowledge-driven machine learning (e.g., expert systems) to data-driven AI. Fueled by the exponential growth of data generated by computing devices and shared on the internet, researchers curated data for the development of machine learning algorithms. One significant effort was the development of the ImageNet...

Table of contents

• Cover
• Half-Title
• Series
• Title
• Copyright
• Dedication
• Contents
• Foreword
• Preface
• List of Contributors
• Editor
• 1 Unraveling Data Science, Artificial Intelligence, and Autonomy
• 2 Unlock the True Power of Data Analytics with Artificial Intelligence
• 3 Machine Intelligence and Managerial Decision-Making
• 4 Measurement Issues in the Uncanny Valley: The Interaction between Artificial Intelligence and Data Analytics
• 5 An Overview of Deep Learning in Industry
• 6 Chinese AI Policy and the Path to Global Leadership: Competition, Protectionism, and Security
• 7 Natural Language Processing in Data Analytics
• 8 AI in Smart Cities Development: A Perspective of Strategic Risk Management
• 9 Predicting Patient Missed Appointments in the Academic Dental Clinic
• 10 Machine Learning in Cognitive Neuroimaging
• 11 People, Competencies, and Capabilities Are Core Elements in Digital Transformation: A Case Study of a Digital Transformation Project at ABB
• 12 AI-Informed Analytics Cycle: Reinforcing Concepts
• Index