Table of Contents

Enum ClassificationTaskType

Namespace
AiDotNet.Enums
Assembly
AiDotNet.dll

Specifies the type of classification task being performed.

public enum ClassificationTaskType

Fields

Binary = 0

Binary classification with exactly two classes (e.g., spam/not-spam, positive/negative).

Binary classification is the simplest form of classification where each sample belongs to exactly one of two classes. The output is typically a single probability value representing the likelihood of the positive class.

For Beginners: Binary classification answers yes/no questions.

Examples:

  • Is this email spam? (Yes/No)
  • Will this customer churn? (Yes/No)
  • Is this transaction fraudulent? (Yes/No)
  • Does this patient have the disease? (Yes/No)

The model outputs a probability between 0 and 1, and you typically use a threshold (usually 0.5) to make the final yes/no decision.

MultiClass = 1

Multi-class classification where each sample belongs to exactly one of multiple classes.

Multi-class classification extends binary classification to multiple mutually exclusive classes. The output is a probability distribution over all classes, and the predicted class is typically the one with the highest probability.

For Beginners: Multi-class classification picks one category from many options.

Examples:

  • What animal is in this image? (Cat, Dog, Bird, Fish, etc.)
  • What genre is this movie? (Action, Comedy, Drama, Horror, etc.)
  • What digit is this? (0, 1, 2, 3, 4, 5, 6, 7, 8, 9)
  • What language is this text? (English, Spanish, French, etc.)

Key difference from multi-label: each sample can only belong to ONE class. A movie can't be both "Comedy" and "Drama" in multi-class classification.

MultiLabel = 2

Multi-label classification where each sample can belong to multiple classes simultaneously.

Multi-label classification allows each sample to be assigned to multiple classes at the same time. The output is a set of binary predictions (one for each possible label), indicating which labels apply to the sample.

For Beginners: Multi-label classification assigns multiple tags to each item.

Examples:

  • What topics are in this article? (Politics, Sports, Technology - multiple can apply)
  • What objects are in this image? (Person, Car, Tree - multiple can appear)
  • What genres does this movie belong to? (Action-Comedy, Drama-Romance)
  • What symptoms does this patient have? (Fever, Cough, Fatigue)

Key difference from multi-class: an item CAN belong to multiple categories at once. A movie can be both "Comedy" AND "Drama" in multi-label classification.

Ordinal = 3

Ordinal classification where classes have a natural ordering.

Ordinal classification is similar to multi-class classification, but the classes have an inherent order. The model should respect this ordering, penalizing predictions that are "far" from the true class more than predictions that are "close."

For Beginners: Ordinal classification predicts categories that have a natural order.

Examples:

  • Customer satisfaction: Poor, Fair, Good, Excellent (ordered from worst to best)
  • Product rating: 1 star, 2 stars, 3 stars, 4 stars, 5 stars
  • Education level: High School, Bachelor's, Master's, PhD
  • Pain level: None, Mild, Moderate, Severe

The key insight is that predicting "Fair" when the answer is "Good" is less wrong than predicting "Poor" - they're both wrong, but one is closer to the truth.

Unlike regular multi-class where all mistakes are equally wrong, ordinal classification considers how far off the prediction is from the correct answer.

Remarks

Classification task types determine how the model interprets the target variable and how predictions are structured. Different task types require different output formats and loss functions.

For Beginners: Classification is about putting things into categories.

Think of it like sorting mail:

  • Binary: Is this spam or not spam? (2 categories)
  • MultiClass: Is this a bill, letter, package, or advertisement? (multiple exclusive categories)
  • MultiLabel: Mark all that apply: urgent, personal, work-related (multiple overlapping labels)
  • Ordinal: Rate satisfaction: Poor, Fair, Good, Excellent (ordered categories)

The task type tells the model what kind of answer you're expecting.

Edit this page