zero-one-software / linear_regression_trend

A library for calculating linear regression against time series data

Linear Regression Trend Calculator

This is a simple gem to help calculate linear regression trendlines for a given list of samples.

Note

This gem was created using this gist. originally authored by Scott Baldwin.. Thanks dude.

Installation

Add this line to your application's Gemfile:

gem 'linear_regression_trend'

And then execute:

$ bundle

Or install it yourself as:

$ gem install linear_regression_trend

Usage: Standard usage

my_array_of_samples = [10, 14, 18, 22, 26]
trender = LinearRegressionTrend::Calculator.new(my_array_of_samples)

trended_samples = trender.trend     # => [array of trend points]
trend_slope     = trender.slope     # => 4
trend_intercept = trender.intercept # => 6

Usage: non-negative trending

Sometimes you want a trend, and that trend happens to be a descending one. Normally this is fine. However there are instances where that trend calculation can go into the negative. Again, sometimes this is fine, but sometimes this doesnt make sense, e.g. packet loss, or latency on a network can never be less than zero (would be cool though). In cases like this, you'd wanna tell the library to stop trending past zero. Observe:

my_array_of_samples = [1,2,3,0,10,18,18,18,1,1,1,1,0.5,0.5,0.5,0.5,05,0.25,0.25,0.1,0.1,0,0,0,0]
trender = LinearRegressionTrend::Calculator.new(my_array_of_samples, non_negative: true)

trended_samples = trender.trend     # => [array of trend points]
trend_slope     = trender.slope     # => a negative number, indicating a descending trend
trended_samples.last == 0           # => true, would've be -0.8927 otherwise

Usage: Forecasting values

Let's assume you're have a value for each day of the month, and it's currently the middle of the month. If you want to know what the trend will look like for the remainder of the month, you can forecast the trend past the samples supplied.

samples = [10, 14, 18, 22, 26, 15, 19, 12, 1, 39, 31, 33, 16, 9, 12] # 15 days of samples
trender = LinearRegressionTrend::Calculator.new(samples)

forecast = trender.forecast_next(15) # => [20.267, 20.492, 20.717, 20.942, 21.167, 21.392, 21.617, 21.842, 22.067, 22.292, 22.517, 22.742, 22.967, 23.192, 23.417]

In this example, the actual trend ends at 20.042, and the forecast continues smoothly using the same slope from that point on.

This functionality will obey the non_negative flag's setting and forecast down to 0 if set.

Development and extra documentation

Check out the tests in the /spec dir, and also install iRuby notebook, and take a look at the live documentation in /doc.

Contributing

1. Fork it ( https://github.com/zero-one-software/linear_regression_trend/fork )
2. Create your feature branch (git checkout -b my-new-feature)
3. Commit your changes (git commit -am 'Add some feature')
4. Push to the branch (git push origin my-new-feature)
5. Add some tests
6. Create a new Pull Request
7. Do a happy dance