Algorithme des k plus proches voisins

New Section

This section introduces the algorithm of the 10-nearest neighbors, discussing the problem, the algorithm's principle, and additional details. The algorithm belongs to machine learning algorithms and is supervised learning.

Problem Introduction

• The algorithm of the 10-nearest neighbors is presented.

• It falls under supervised learning where data considered are labeled into classes.

• Data points are represented in an orthogonal coordinate system based on their characteristics like size and weight.

Algorithm Principle

• The goal is to assign a class (e.g., stars or triangles) to a new point based on its nearest neighbors.

• Initially, the closest neighbor determines the class assigned.

• Expanding to consider 5 and more nearest neighbors for classification.

Additional Insights

• Determining the most frequent class among the nearest neighbors for classification.

• Importance of parameter selection such as k value and distance metric choice like Manhattan or Minkowski distances.

New Section

This part elaborates on further aspects of the algorithm, including dataset considerations, distance functions, and parameter tuning for optimal performance.

Dataset Considerations

• Dataset consists of labeled data points with multiple features.

• A distance function calculates distances between data points for classification purposes.

Distance Functions

• Utilizing Euclidean distance for calculating distances in an orthogonal coordinate system.

• Mention of alternative distance metrics like Manhattan or Chebyshev distances.

Parameter Tuning

• Selecting an appropriate k value through iterative testing for optimal performance.