Top Page | Upper Page | Contents | About This Site | JAPANESE

Neighboring method

The neighbor method is a method of trying to achieve the purpose with only a small part of the data within a certain range of data.

In daily life, when making a guess, we sometimes refer to "recent trends" and "peripheral trends", but the idea of ??the neighbor method uses this idea.

Pros and cons

The neighbor method is an advantageous method for local guessing because it uses only some data, not the entire data. Even if you can't express the relationship between Y and X with a mathematical formula, you can make a reliable guess. The weaknesses are the backfires of the strengths. Since you only see some data, it is very sensitive to nearby data.

Also, since the entire data is not used, it is not easily affected by outliers . It can be said to be one of the robust analyzes .

Method using neighbor method algorithm

• k-nearest neighbor method :Guess the value from the value in the neighborhood Used as Label Classification and Regression Analysis.
• LOF :Estimate the degree of isolation from the density (how data is collected) in the vicinity. Used as Outlier Detection.
• One-class Minimum distance method :Estimate the degree of isolation from the distance in the vicinity. Used as One-Class Model.

NEXT k-NN