NAME
GitInsight - Predict your github contributions using Bayesian inference
and Markov chain
SYNOPSIS
gitinsight --username [githubusername] (--nodaystats) (--accuracy) #using the shipped bin
#or using the module
my $Insight= GitInsight->new(no_day_stats => 0, username => "markov", accuracy=> 1);
my $Result= $Insight->process;
my $accuracy = $Insight->{accuracy};
$Result = $Insight->{result};
# $Result contains the next week predictions and is an arrayref of arrayrefs [ [ 'Sat', 1, '2014-07-1', [ 0 , '0.151515151515152', '0.0606060606060606', '0.0404040404040404', 0 ] ], .. [ 'DayofWeek', 'winner_label', 'day' , [ probability_label_0, probability_label_1, probability_label_2, probability_label_3, probability_label_4 ] ]]
DESCRIPTION
GitInsight is module that allow you to predict your github contributions
in the "calendar contribution" style of github (the table of
contribution that you see on your profile page).
HOW DOES IT WORK?
GitInsight generates a transation probrability matrix from your github
contrib_calendar to compute the possibles states for the following days.
Given that GitHub split the states thru 5 states (or here also called
label), the probability can be inferenced by using Bayesian methods to
update the beliefs of the possible state transition, while markov chain
is used to predict the states. The output of the submitted data is then
plotted using Cellular Automata.
THEORY
We trace the transitions states in a matrix and increasing the count as
far as we observe a transition
(), then we inference
the probabilities using Bayesan method
.
INSTALLATION
GitInsight requires the installation of gsl (GNU scientific library),
gd(http://libgd.org/), PDL and PDL::Stats (to be installed after the gsl
library set).
on Debian:
apt-get install gsl-bin libgs10-devt
apt-get install pdl libpdl-stats-perl libgd2-xpm-dev
It's reccomended to use cpanm to install all the required deps, install
it thru your package manager or just do:
cpan App::cpanminus
After the installation of gsl, clone the repository and install all the
dependencies with cpanm:
cpanm --installdeps .
Then install it as usual:
perl Build.PL
./Build
./Build test #ensure that the module works correctly
./Build install
OPTIONS
username
required, it's the GitHub username used to calculate the prediction
ca_output
you can enable/disable the cellular autmata output using this option
(1/0)
no_day_stats
setting this option to 1, will slightly change the prediction: it will
be calculated a unique transition matrix instead one for each day
left_cutoff
used to cut the days from the start (e.g. if you want to delete the
first 20 days from the prediction, just set this to 20)
cutoff_offset
used to select a range where the prediction happens (e.g. if you want to
calculate the prediction of a portion of your year of contribution)
file_output
here you can choose the file output name for ca.
accuracy
Enable/disable accuracy calculation (1/0)
verbose
Enable/disable verbosity (1/0)
METHODS
contrib_calendar($username)
Fetches the github contrib_calendar of the specified user
process
Calculate the predictions and generate the CA
start_day
Returns the first day of the contrib_calendar
last_day
Returns the last day of the contrib calendar (prediction included)
prediction_start_day
Returns the first day of the prediction (7 days of predictions)
AUTHOR
mudler
COPYRIGHT
Copyright 2014- mudler
LICENSE
This library is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.
SEE ALSO
GitInsight::Util, PDL, PDL::Stats