liveBook · Manning

Dataset

As explained in Grokking Machine Learning, Second Edition

A dataset is a collection of related data records organized for analysis or computational processing.

A dataset is organized information: a collection of related records arranged for analysis or computational processing. In a table, each row is a data point, and features describe that point. A dataset of pets, for example, uses one row for each pet and columns for its features. section 2.2

Datasets support several basic machine-learning activities:

  • Plot records to look for patterns, such as the relationship between a house’s number of rooms and its price. section 3.3
  • Separate features from labels, such as word counts from happy-or-sad sentence outcomes. section 5.18
  • Compare predictions with known labels, even when a model cannot classify every point correctly. section 5.6

A small house dataset

A table can pair the number of rooms with a house price: 1 room and 155, 2 rooms and 197, 3 rooms and 244, 5 rooms and 356, and 6 rooms and 407. The missing price for four rooms creates a simple prediction problem. section 3.3

Table 3.2 A slightly more complicated dataset of houses with their number of rooms and their price
Number of rooms
Price
1
155
2
197
3
244
4
?
5
356
6
407
7
448

Use the table to inspect how rows pair house features with prices and how one missing value fits into the dataset. section 3.3

How Grokking Machine Learning, Second Edition covers it

How deep it goes
Develops datasets from tables to plots, NumPy arrays, labels, predictions, and evaluation, without pursuing data collection or cleaning.
Examples
Uses house prices, alien-language sentiment, coronavirus diagnoses, and spam-or-ham email records.
What it assumes
Assumes only an understanding of information and tables, then introduces rows, data points, and features.
What you can do after
Apply datasets by plotting records, representing features and labels in NumPy arrays, and evaluating predictions.
How it connects
Connects data with features, labels, linear regression, perceptrons, logistic classifiers, and classification metrics.

Also explained in

Compare how 30 books explain dataset →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Grokking Machine Learning, Second Edition. The section links point to the parts of the book it draws on.