Samenvatting DTZ: Lectures
Lecture 1: Introduction to Data Science in Healthcare
This is an introductory lecture of the course. The lecture will start with the general
motivation for the course, a set of initial key concepts that we have to familiarize with
during the course. We progress with application examples where data science has been
helping caregivers to provide better patient care. We continue with an overview of the data
science lifecycle and explain how we structure the course in relation to it. In the final part
of this initial lecture will explain once more the key rules and the structure for a productive
course.
Lecture 2: Object Oriented Programming
This lecture will focus on the concerts of Object Oriented Programming. In the lecture you
will be introduced to the concepts of Class and Objects and how data types, data structures
and control flows are used together to build an object oriented program. The lecture has
the following learning objectives: Understanding of object oriented programming concepts
and the understanding of the main properties of object oriented programming.
Lecture 3: Introduction to Machine Learning
The lecture gives a fundamental introduction AI and its relationship with ML, elaborates on
input and output knowledge and different types of learning. The learning objectives
include: (i) basic principles of machine learning, (ii) identify a machine learning problem,
Lecture 4: Simple machine learning algorithms
This lecture builds upon the previous week's lecture with further knowledge of basic
machine learning algorithms. The learning objectives include: (i) understanding of basic
machine learning algorithms, and (ii) application appropriate basic machine learning
techniques in biomedical science.
Lecture 5: Advanced machine learning algorithms
This lecture first gives an in-depth introduction to two approaches to clustering and
classification: Gaussian mixture models and artificial neural networks. We review the
mathematical underpinnings of both approaches, contextualize them in relation to previous
algorithms students have encountered (e.g., K-means clustering, logistic regression), and
reflect on their limitations and appropriate use-cases.
Lecture 6: Data Visualization
This lecture will be centered on data Visualization and its importance in both the data
exploration phase as well as in validation and the communication of data analysis results.
The lecture has the following learning objectives: Understanding the methods for
representing data structures; mastering the different visualization techniques and
understanding how to effectively communicate findings.
Lecture 7: Text analysis
This lecture will discuss how unstructured data formats, such as text, can be analyzed.
This lecture will focus on introducing some simple text analysis techniques. The lecture has
the following learning objectives: Understanding of how to represent unstructured text
documents with appropriate format and the acquisition of knowledge over text mining
algorithms.
Openingslecture
Data science is an interdisciplinary field that uses scientific methods,
processes, algorithms and systems to extract knowledge and insights from
data
,→ humans are now able to make decisions based on structured information
Patient Data Space
Patients have multiple data points spread across various institutions,
private companies and doctor offices, if data are analysed they can
provide a glimpse into the needs of patients in the short and long term.
Data Science for genetics: studying the role that genetics plays in disease
across populations. (example: polygenic risk score informs people about
their risk of developing a disease)
The Data Science Lifecycle
Step 1: Frame the problem and formulate the research question
Step 2: Obtain your data
Step 3: Scrub your data
Step 4: Explore your data
Step 5-6: Model your data
Step 7: Communicate and interpret results of the analysis.
We typically use data science to answer 5 types of questions:
- How much or how many? (regression)
- Which category? (classification)
- Which group? (clustering)
- Is this weird? (anomaly detection)
- Which option should be taken? (recommendation)
Data is split into two parts:
- Training set → used to build the algorithm
, - Test set → used to validate how good the algorithm is
Classification Task
Imagine that you want to predict mortality for Heart Failure (HF) patients.
- You have several data collected for this analysis. These data include
demographics, ecography, education ecc (these data are called
features).
- You know from the data if the patient passed away or not (the label).
In general, the label can be of two or more classes
- The classification task will assign a probability of being alive or dead
based on the information.
- When the model learns to classify the HF patients, you can use new
data to make a prediction
Regression Task
When the output is a continuous value, the task is a regression.
- For instance, following the example on heart failure patients, you
may need to forecast the healthcare expense generated by HF
patients
- You select a range of features like health status, previous use of
healthcare services, hospitalization rate.
- The system is trained to estimate the healthcare expense with the
lowest possible error
Linear regression: finds a way to correlate each feature to the output to
help predict future values → Regression
Logistic regression: extension of linear regression that’s used for
classification tasks. The output variable is binary (e.g., only black or white)
rather than continuous (e.g., an infinite list of potential colors) →
Classification
Overfitting: too much reliance on training data
, High variance: model changes significantly based on training data
Underfitting: a failure to learn the relationships in the training data
High bias: assumptions about model lead to ignoring training data
- Overfitting and underfitting cause poor generation on the test set
- A validation set for model tuning can prevent under and overfitting
Numbers
Python distinguishes between two different types of numbers:
- Integers are called int values in the Python language.
They can only represent whole numbers (negative, zero, or positive)
Do not have a fractional component
- Real numbers are called float values (or floating point values).
- The type of a number is evident from the way it is displayed: float
values always have a decimal point.
Strings
A piece of text represented in a computer is called a string
- A string can represent a word, a sentence, or contents of every book
in a library
- From an existing string, related string can be constructed using
function that operate on strings
- These functions are called by placing a dot after the string, then
calling the function.
Lists
- The list is a most versatile datatype available in Python
- It is written as a list of comma-separated values (items) between
square brackets
Lecture 1: Introduction to Data Science in Healthcare
This is an introductory lecture of the course. The lecture will start with the general
motivation for the course, a set of initial key concepts that we have to familiarize with
during the course. We progress with application examples where data science has been
helping caregivers to provide better patient care. We continue with an overview of the data
science lifecycle and explain how we structure the course in relation to it. In the final part
of this initial lecture will explain once more the key rules and the structure for a productive
course.
Lecture 2: Object Oriented Programming
This lecture will focus on the concerts of Object Oriented Programming. In the lecture you
will be introduced to the concepts of Class and Objects and how data types, data structures
and control flows are used together to build an object oriented program. The lecture has
the following learning objectives: Understanding of object oriented programming concepts
and the understanding of the main properties of object oriented programming.
Lecture 3: Introduction to Machine Learning
The lecture gives a fundamental introduction AI and its relationship with ML, elaborates on
input and output knowledge and different types of learning. The learning objectives
include: (i) basic principles of machine learning, (ii) identify a machine learning problem,
Lecture 4: Simple machine learning algorithms
This lecture builds upon the previous week's lecture with further knowledge of basic
machine learning algorithms. The learning objectives include: (i) understanding of basic
machine learning algorithms, and (ii) application appropriate basic machine learning
techniques in biomedical science.
Lecture 5: Advanced machine learning algorithms
This lecture first gives an in-depth introduction to two approaches to clustering and
classification: Gaussian mixture models and artificial neural networks. We review the
mathematical underpinnings of both approaches, contextualize them in relation to previous
algorithms students have encountered (e.g., K-means clustering, logistic regression), and
reflect on their limitations and appropriate use-cases.
Lecture 6: Data Visualization
This lecture will be centered on data Visualization and its importance in both the data
exploration phase as well as in validation and the communication of data analysis results.
The lecture has the following learning objectives: Understanding the methods for
representing data structures; mastering the different visualization techniques and
understanding how to effectively communicate findings.
Lecture 7: Text analysis
This lecture will discuss how unstructured data formats, such as text, can be analyzed.
This lecture will focus on introducing some simple text analysis techniques. The lecture has
the following learning objectives: Understanding of how to represent unstructured text
documents with appropriate format and the acquisition of knowledge over text mining
algorithms.
Openingslecture
Data science is an interdisciplinary field that uses scientific methods,
processes, algorithms and systems to extract knowledge and insights from
data
,→ humans are now able to make decisions based on structured information
Patient Data Space
Patients have multiple data points spread across various institutions,
private companies and doctor offices, if data are analysed they can
provide a glimpse into the needs of patients in the short and long term.
Data Science for genetics: studying the role that genetics plays in disease
across populations. (example: polygenic risk score informs people about
their risk of developing a disease)
The Data Science Lifecycle
Step 1: Frame the problem and formulate the research question
Step 2: Obtain your data
Step 3: Scrub your data
Step 4: Explore your data
Step 5-6: Model your data
Step 7: Communicate and interpret results of the analysis.
We typically use data science to answer 5 types of questions:
- How much or how many? (regression)
- Which category? (classification)
- Which group? (clustering)
- Is this weird? (anomaly detection)
- Which option should be taken? (recommendation)
Data is split into two parts:
- Training set → used to build the algorithm
, - Test set → used to validate how good the algorithm is
Classification Task
Imagine that you want to predict mortality for Heart Failure (HF) patients.
- You have several data collected for this analysis. These data include
demographics, ecography, education ecc (these data are called
features).
- You know from the data if the patient passed away or not (the label).
In general, the label can be of two or more classes
- The classification task will assign a probability of being alive or dead
based on the information.
- When the model learns to classify the HF patients, you can use new
data to make a prediction
Regression Task
When the output is a continuous value, the task is a regression.
- For instance, following the example on heart failure patients, you
may need to forecast the healthcare expense generated by HF
patients
- You select a range of features like health status, previous use of
healthcare services, hospitalization rate.
- The system is trained to estimate the healthcare expense with the
lowest possible error
Linear regression: finds a way to correlate each feature to the output to
help predict future values → Regression
Logistic regression: extension of linear regression that’s used for
classification tasks. The output variable is binary (e.g., only black or white)
rather than continuous (e.g., an infinite list of potential colors) →
Classification
Overfitting: too much reliance on training data
, High variance: model changes significantly based on training data
Underfitting: a failure to learn the relationships in the training data
High bias: assumptions about model lead to ignoring training data
- Overfitting and underfitting cause poor generation on the test set
- A validation set for model tuning can prevent under and overfitting
Numbers
Python distinguishes between two different types of numbers:
- Integers are called int values in the Python language.
They can only represent whole numbers (negative, zero, or positive)
Do not have a fractional component
- Real numbers are called float values (or floating point values).
- The type of a number is evident from the way it is displayed: float
values always have a decimal point.
Strings
A piece of text represented in a computer is called a string
- A string can represent a word, a sentence, or contents of every book
in a library
- From an existing string, related string can be constructed using
function that operate on strings
- These functions are called by placing a dot after the string, then
calling the function.
Lists
- The list is a most versatile datatype available in Python
- It is written as a list of comma-separated values (items) between
square brackets
