Artificial Intelligence and Data Analytics (AIDA) Guidebook

Figure 4: Comparison of Symbolic and Statistical Modeling

While these early systems were developed and altered manually, today's AI revolution is driven
by machine learning algorithms – statistical models that are automatically refined by repeatedly
executing thousands of data sets through an algorithm that improves itself or learns to improve
its outcomes. Advances in modern AI systems
are all made possible by the confluence of four AI Example: Smart Assistants
converging technological trends: increases in
computing speed, addressable storage Smart assistants are devices loaded with
capacity, network bandwidth, and advanced software that you use to access information,
mathematics. We can access larger and more perform tasks or control other devices. You
comprehensive data sets, as math gets more
can have a smart assistant on your computer
efficient, trustworthy, and robust in its
or your phone, but most people use them
representation of reality. This allows the
through a smart speaker. Well-known brands
training of more models faster and the pursuit
include Alexa, Siri, Cortana and Google
of AI in a growing range of applications is now
feasible because of advances in computational Assistant.
complexity.
In addition to symbolic modeling of AI and machine learning, there is a complementary
perspective called human-machine teaming or human-machine learning that evolved from
intelligence augmentation, which is the use of technology to enhance human intelligence and
decision making. Although intelligence augmentation evolved around the same time as AI and
machine learning, this perspective is relevant more than ever as we seek to leverage the
advantages gained by human intelligence and machine learning models.
Development of modern AI systems focuses on continuously developing and deploying updates,
often referred to as Agile Development. Some of the modern approaches to AI system
development include:

Page 12
Artificial Intelligence and Data Analytics (AIDA) Guidebook

• DevOps - the combination of software and hardware systems aimed at continuously

developing and deploying software to increase operational efficiencies.

• DevSecOps - the integration of security requirements and testing within the DevOps
process.

• AIOps - use of AI in supporting network operations, including anomaly and threat

detection.

• MLOps - applying the continuous development and integration approach to the

development of machine learning solutions, requiring additional automation step of
automating training data access and preparation.
This guidebook recommends implementing MLOps to develop AI systems whenever feasible
and Section 0 will provide additional detail on the Machine Learning Pipeline methodology and
its current applications.
The remainder of the guidebook describes key AI and data analytics concepts and
considerations relevant to the implementation of these models and analytics tools. Figure 5
details the primary intents of this guidebook and how each section of the document relates
back to those intents.

Page 13
Artificial Intelligence and Data Analytics (AIDA) Guidebook

Figure 5: Intent and Content of AIDA Guidebook

Page 14
Artificial Intelligence and Data Analytics (AIDA) Guidebook

5 Machine Learning Methodology

This section provides an overview of how developers effectively build, evaluate, and manage
analytic and learning systems through a machine-learning pipeline. A machine-learning pipeline
is a way to codify and automate the workflow necessary to produce a machine-learning model.
Machine learning pipelines consist of multiple sequential steps that do everything from data
extraction and preprocessing to model training and deployment. Figure 6 provides a high-level
example of a machine-learning pipeline, with each step described in more detail to follow.

Figure 6: Machine Learning Pipeline

Pipeline steps:

1. Establish AI system goal – traditional goals of AI research include reasoning, knowledge

representation, planning, learning, natural language processing, perception, and the
ability to move and manipulate objects.
2. Establish requirements – consider desired performance, usability, integration, and
statistical behavior.
3. Identify AI solution design – identify the algorithms and programming language to be
used.
4. Identify use constraints – constraints enumerate the possible values a set of variables
may take in a given world.
5. Identify required data sets – the more complex your model becomes, the more data you
will need to determine its parameters.

Page 15