This document only contains the Key Concepts.
Buy my other summary for a 50+ pager for a more comprehensive explanation of everything
https://www.stuvia.com/doc/1809995/data-science-summary-key-concepts-more-compact-summary
, Key concepts
List of steps to take in data science Execute experiment:
1. Explore 1. Task definition
2. Formulate research question 2. Data collection
3. Data exploration
3. Structure and annotate data
4. Preprocessing
4. Develop and apply learning 5. Model learning
techniques 6. Evaluation
5. Evaluate on data
6. Answer the research question
List three challenges of working with data:
1. Noisy data
2. Small data / large data
3. Data can be incomplete
different sampling rates, different formats, wrongly chosen or irrelevant variables, large / unknown
number of classes, class imbalance, heterogeneous data / features, new domain, …
How to give a clear definition of a task, based on a given data set
:
● Research question
● Determine supervised vs unsupervised
● Classification or regression (or clustering if its unsu pervised)
● Problem definition:
○ Features and their type (binary, nominal(multi categorical), numerical)
○ Target labels and their type (binary, nominal, numerical)
Use median vs mean: Mean when the distribution is symmetrical and median otherwise.
Explain simple linear regression, multiple linear regression and logistic regression:
● Linear regression: Defines the relationship between two variables.
used to handle basic regressions (when the relation between two vars is clear
and simple),
● Multiple linear regression: defines relationship by more than one value
Used more complex connections between data (house prices need more
variables than bedrooms for example)
● Logistic regression: Discriminative model that learns to distinguish between two
classes
Used to handle classification problems
Buy my other summary for a 50+ pager for a more comprehensive explanation of everything
https://www.stuvia.com/doc/1809995/data-science-summary-key-concepts-more-compact-summary
, Key concepts
List of steps to take in data science Execute experiment:
1. Explore 1. Task definition
2. Formulate research question 2. Data collection
3. Data exploration
3. Structure and annotate data
4. Preprocessing
4. Develop and apply learning 5. Model learning
techniques 6. Evaluation
5. Evaluate on data
6. Answer the research question
List three challenges of working with data:
1. Noisy data
2. Small data / large data
3. Data can be incomplete
different sampling rates, different formats, wrongly chosen or irrelevant variables, large / unknown
number of classes, class imbalance, heterogeneous data / features, new domain, …
How to give a clear definition of a task, based on a given data set
:
● Research question
● Determine supervised vs unsupervised
● Classification or regression (or clustering if its unsu pervised)
● Problem definition:
○ Features and their type (binary, nominal(multi categorical), numerical)
○ Target labels and their type (binary, nominal, numerical)
Use median vs mean: Mean when the distribution is symmetrical and median otherwise.
Explain simple linear regression, multiple linear regression and logistic regression:
● Linear regression: Defines the relationship between two variables.
used to handle basic regressions (when the relation between two vars is clear
and simple),
● Multiple linear regression: defines relationship by more than one value
Used more complex connections between data (house prices need more
variables than bedrooms for example)
● Logistic regression: Discriminative model that learns to distinguish between two
classes
Used to handle classification problems
