13 Smart Learning Using Big and Small Data for Mobile and IOT e-Health
611
13.1.2 Incorporating Domain Knowledge in Data-Driven
Learning
To overcome these challenges, incorporating domain knowledge in the data-driven
learning framework would help. First, structural knowledge bases, such as decision
rules used in clinical practice guidelines (CPG), can help shorten the development
cycle and reduce the amount of data needed for training. Second, it can help avoid
the biased results because of its capability to better generalize and make the results
more self-interpretable among domain experts.
However, how to leverage structural knowledge bases to guide data-driven
learning remains to be a challenging issue. This is especially true in healthcare,
which has amassed evidence-based, comprehensive knowledge bases over the
years and is in need for interpretable decision rules to trust AI/ML-augmented
recommendations [50–52].
Previous solutions could be roughly divided into two categories. One is the
homogeneous method, which integrates knowledge-based representations (e.g.,
logic rules) with data-driven learning tasks into a unified inference framework
[53–55]. The other is the heterogeneous method, which integrates the respective
inference results from knowledge-based and data-driven modules [56]. While both
categories of methods have its merits, each has its own gap to fill before knowledge
bases and data-driven models can work together to generate real-world evidence in
healthcare practice.
13.1.3 Structure of the Book Chapter
In the rest of this book chapter, we will use three case studies to illustrate the
challenges and opportunities facing the development of a smart learning framework.
Across the three case studies, we will especially focus on the lessons learned from
applying smart learning for mobile and IOT e-health and embedding actionable
insights in care flows for decision-making.
The aim of this chapter is twofold. First, we would like to provide an overview
of the status quo of this emerging field of applying and developing a smart learning
framework for mobile and IOT e-health technologies. Second, we would identify
gaps and gather requirements to achieve the goal of generating real-world evidence
in practice. After introducing the cases studies, we will discuss the best practices
and barriers of implementing such a smart learning framework in real-life settings
and further developing a hybrid infrastructure that will allow the generation of
knowledge-augmented, data-derived insights for mobile and IOT e-health.
611
13.1.2 Incorporating Domain Knowledge in Data-Driven
Learning
To overcome these challenges, incorporating domain knowledge in the data-driven
learning framework would help. First, structural knowledge bases, such as decision
rules used in clinical practice guidelines (CPG), can help shorten the development
cycle and reduce the amount of data needed for training. Second, it can help avoid
the biased results because of its capability to better generalize and make the results
more self-interpretable among domain experts.
However, how to leverage structural knowledge bases to guide data-driven
learning remains to be a challenging issue. This is especially true in healthcare,
which has amassed evidence-based, comprehensive knowledge bases over the
years and is in need for interpretable decision rules to trust AI/ML-augmented
recommendations [50–52].
Previous solutions could be roughly divided into two categories. One is the
homogeneous method, which integrates knowledge-based representations (e.g.,
logic rules) with data-driven learning tasks into a unified inference framework
[53–55]. The other is the heterogeneous method, which integrates the respective
inference results from knowledge-based and data-driven modules [56]. While both
categories of methods have its merits, each has its own gap to fill before knowledge
bases and data-driven models can work together to generate real-world evidence in
healthcare practice.
13.1.3 Structure of the Book Chapter
In the rest of this book chapter, we will use three case studies to illustrate the
challenges and opportunities facing the development of a smart learning framework.
Across the three case studies, we will especially focus on the lessons learned from
applying smart learning for mobile and IOT e-health and embedding actionable
insights in care flows for decision-making.
The aim of this chapter is twofold. First, we would like to provide an overview
of the status quo of this emerging field of applying and developing a smart learning
framework for mobile and IOT e-health technologies. Second, we would identify
gaps and gather requirements to achieve the goal of generating real-world evidence
in practice. After introducing the cases studies, we will discuss the best practices
and barriers of implementing such a smart learning framework in real-life settings
and further developing a hybrid infrastructure that will allow the generation of
knowledge-augmented, data-derived insights for mobile and IOT e-health.