Lecture 3: Data Exploration and Discretization

Flashcards covering data exploration concepts (summary statistics, visualization) and discretization methods from Lecture 3.

Summary statistics

A set of measures that summarize data, e.g., frequency and mean.

Frequency

The percentage of times a value occurs in the data set.

Mode

The most frequent attribute value.

Mean

Arithmetic average; a location measure that is sensitive to outliers.

Median

The middle value; a measure of central tendency often used as an alternative to the mean.

Range

Difference between the maximum and minimum values.

Variance

A measure of the spread of a data set; a common dispersion metric.

Visualization

Conversion of data into visual representations to reveal patterns, relationships, and outliers.

Scatter plot

A two-dimensional plot showing relationships between two numeric attributes; can use size, shape, and color to encode extra attributes.

Histogram

A chart showing the distribution of a single variable by binning values into intervals.

Discretization

Turning a numeric (continuous) attribute into a categorical attribute by dividing its range into sub-ranges (bins).

Bin (bucket)

A sub-range of values used in discretization.

Equal-width discretization

Divides the value range into N equal-sized subranges; bin width = (max – min) / N.

Equal-frequency discretization

Divides the range into N bins so each bin holds roughly the same number of instances.

Unsupervised discretization

Discretization methods that do not use class values when creating bins (e.g., equal-width, equal-frequency).

Supervised discretization

Discretization methods that consider class values to choose bin boundaries.

Entropy-based discretization

A supervised method using information entropy to select bin boundaries for better class separation.

Iris dataset

A classic data set with three flower classes (Setosa, Virginica, Versicolor) and four attributes (sepal/petal length/width).