C1

Question 1

question-driven research

Answer 1

start with research question:

1. formulate hypothesis
2. design experiment
3. collect required data
4. analyse data
5. accept or reject hypothesis

Question 2

data-driven research

Answer 2

start with data:

1. explore the data
2. formulate research question
3. structure and label the data
4. develop, apply and optimize learning techniques
5. evaluate
6. answer research question

Question 3

data science experiment steps

Answer 3

1. task definition
2. data sampling (and labelling)
3. data exploration
4. pre-processing (and feature extraction)
5. model learning
6. evaluation

Question 4

structured vs unstructured data

Answer 4

• structured data: a database with tables (rows and columns) and a relational key (logical expression of information)
• semi-structured data: NoSQL, XML, JSON, CSV
• structured data: anything else (image data, text data, sound data, sensor data, possibly with labels)

Question 5

data science techniques

Answer 5

• data exploration
• data mining / data exploration (unsupervised)
• machine learning (supervised)

Question 6

logistic regression

Answer 6

discriminative model that learns to distinguish between two classes, minimizing the loss funcion with gradient descent and transforming the output to a probability with the sigmoid function

Question 7

data challenges

Answer 7

noisy data, large data, small data, unknown classes, class imbalance, heterogeneous data