Introduction to Data Science

Study Notes

Data Science Introduction

Different forms of datasets exist: Record data, Graph and networks, Ordered data.
Record data includes relational tables, data matrix, transaction data, and document data.
Relational tables are highly structured, using columns and rows for efficient data storage.
Data matrix encompasses numerical matrices and crosstabs, presenting data in table form.
Transaction data captures detailed individual transactions, often used for analysis and understanding customer behaviour.
Document data utilizes term-frequency vectors (matrix) to analyze textual content, identifying key terms and relationships.
Entity refers to a specific object or thing.
Attribute is a measurable or observable characteristic of an entity.
Data is the measured value of an attribute.
Data Categorization classifies data into quantitative and qualitative, each having distinct properties.
NOIR topology categorizes data based on its properties - Nominal, Ordinal, Interval, Ratio.
Nominal data is categorized without any inherent order, represented by labels like gender or blood type.
Ordinal data has an order, allowing comparison of categories, for example, ranking of students.
Interval data has order and equal intervals between measurements, allowing for addition and subtraction.
Ratio data is similar to interval data but has a true zero point, allowing for multiplication and division.
Binary Scale is a special case of nominal scale, with only two mutually exclusive categories without inherent ordering.
Symmetric binary variables can be reversed, like "on/off" or "true/false".
Asymmetric binary variables are not reversible, like "positive/negative" or "male/female".
Nominal scale utilizes labels for categorization without inherent order, using letters, numbers, or strings.
Binary scale is a special case of nominal scale, with only two categories.
Ternary scale has three categories, while others have more than two.
Nominal data, even when represented numerically, lacks mathematical interpretation.
Labels in nominal data can be combined to create new nominal variables.
The number of categories in nominal data should be countably finite.