Table of Contents

Interface IFeatureSelector<T, TInput>

Namespace
AiDotNet.Interfaces
Assembly
AiDotNet.dll

Defines an interface for selecting the most relevant features from a dataset.

public interface IFeatureSelector<T, TInput>

Type Parameters

T

The numeric type used for calculations (e.g., double, float).

TInput

Remarks

For Beginners: Feature selection is like choosing which ingredients actually matter for your recipe.

In machine learning, "features" are the individual measurable properties of what you're observing. For example, if you're predicting house prices, features might include:

  • Square footage
  • Number of bedrooms
  • Location
  • Year built
  • etc.

However, not all features are equally useful. Some might:

  • Be redundant (e.g., house area in square feet and square meters)
  • Have little impact on what you're predicting
  • Add "noise" that makes predictions less accurate

Feature selection helps identify which features are most important, which:

  • Makes your model simpler and faster
  • Often improves prediction accuracy
  • Makes results easier to interpret
  • Reduces the risk of "overfitting" (when a model learns noise instead of true patterns)

This interface provides a standard way to implement different feature selection techniques.

Methods

SelectFeatures(TInput)

Selects the most relevant features from the provided feature matrix.

TInput SelectFeatures(TInput allFeaturesMatrix)

Parameters

allFeaturesMatrix TInput

The complete matrix of features where each row represents a sample and each column represents a feature.

Returns

TInput

A matrix containing only the selected features, with the same number of rows but potentially fewer columns.

Remarks

For Beginners: This method analyzes your data and keeps only the columns (features) that matter most.

Imagine you have a spreadsheet with 100 columns of data about houses, but only 10 of those columns actually help predict the house price. This method would identify and keep just those 10 important columns, discarding the rest.

The input parameter:

  • allFeaturesMatrix: Your complete dataset organized as a matrix (like a table or spreadsheet)
    • Each row is one example (like one house)
    • Each column is one feature (like square footage, number of bedrooms, etc.)

The output:

  • A new matrix with the same number of rows (same examples)
  • But fewer columns (only the important features)

Different implementations of this interface might use different techniques to decide which features are important, such as:

  • Statistical tests
  • Correlation with the target variable
  • Machine learning algorithms that can rank feature importance

By removing less important features, your models can run faster and often make better predictions.

Edit this page