pearson's Introduction

Pearson

Pearson correlation coefficient calculator

Installation

You can install the pearson gem with rubygems:

gem install pearson

If you are using Bundler, you can include it into the Gemfile:

gem 'pearson', '~> 1.0'

Usage

scores = {
  'Jack' => {
    'The Godfather' => 2.5,
    'Gattaca' => 3.5,
    'Matrix' => 3.0,
    'American History X' => 3.5,
    'Back to the future' => 2.5
  },
  'Lisa' => {
    'The Godfather' => 1.5,
    'Gattaca' => 2.5,
    'Matrix' => 1.5,
    'Back to the future' => 5.0
  },
  'Tom' => {
    'The Godfather' => 3.5,
    'Gattaca' => 3.0,
    'American History X' => 1.5,
    'Back to the future' => 5.0
  },
  'Sarah' => {
    'The Godfather' => 3.0,
    'Gattaca' => 3.5,
    'Matrix' => 1.5,
    'American History X' => 5.0,
    'Back to the future' => 1.0
  },
  'Mike' => {
    'The Godfather' => 3.0,
    'Back to the future' => 4.0
  }
}

Example 1

Calculate the pearson correlation coefficient between two entities:

Pearson.coefficient(scores, 'Sarah', 'Lisa')
#=> -0.5297608986976613

Example 2

Calculate the closest entity from a given entity:

Pearson.closest_entities(scores, 'Jack', limit: 1)
#=> [['Sarah', 0.7010740719559754]]

Example 3

Calculate the best recommendations for an entity. These recommendations are based on the pearson correlation coefficient. The results are sorted by item ranking:

Pearson.recommendations(scores, 'Mike')
#=> [["Gattaca", 2.75], ["American History X", 1.5], ["Matrix", 1.5]]

pearson's People

Contributors

Stargazers

Watchers

pearson's Issues

Correlation magnitude

If I understand properly... each Pearson coefficient determinates how similar are two entities but you shouldn't care about the 'direction' of the similarity (as least on this 'problem').

https://github.com/alfonsojimenez/pearson/blob/master/lib/pearson.rb#L71 here you discriminate non-positive coefficients but I think you should get better results taking in care the absolute value (the same when you short to get the closest entity).

Recommend Projects