Recommender Systems with Ruby (adding machine learning, statistics, etc)

Ruby in the world of
recommendations
(also machine learning, statistics and visualizations..)
Marcel Caraciolo
@marcelcaraciolo
Developer, Cientist, contributor to the Crab recsys project,
works with Python for 6 years, interested at mobile,
education, machine learning and dataaaaa!
Recife, Brazil - http://aimotion.blogspot.com
Saturday, September 14, 2013

FAÇA BACKUP!

NUNCA:
find
.
-‐type
f
-‐not
-‐name
'*pyc'
|
xargs
rm

Scientific Environment
Presentation & Visualization
Experimentation
(Re-Design)
Data AcquisitionData Analysis

Where is Ruby?
Presentation & Visualization
Experimentation
(Re-Design)
Data AcquisitionData Analysis

Where is Ruby?
Python launched at 1991; Ruby
launched at 1995
Python was highly addopted and
promoted by most of the research and
development team of Google

Where is Ruby?
Python lançado em 1991; Ruby lançado em 1995
Python foi altamente popularizado com a adoção oficial de
boa parte do time de pesquisa do Google
Python has been an important
key of Google since its beginning,
and still continues as our infra-
structure grows, we are always
looking for more people with
skills in this language.
Peter Norvig, Google, Inc.

Where is Ruby?
Python was famous even at some old
scientific articles

Where is Ruby?
Ruby’s popularity exploded at 2004.
Focus on web
Django - 2005; Numpy - 2005;
BioPython - 2001; SAGE - 2005;
Matplotlib- 2000;
Python

Where is Ruby?
Programming comes second to researchers, not
first like us. - “Ruby developer answer”
Python
[(x, x*x) for x in [1,2,3,4] if x != 3]
vs Ruby
`[1,2,3,4].map { |x| [x, x*x] if x != 3 }`
vs Result
[(1,1), (2,4), (4,16)]

Where is Ruby?
Ruby
Python

Hey, Ruby has options!

:(

:D

gem install nmatrix
git clone https://github.com/SciRuby/nmatrix.git
cd nmatrix/
bundle install
rake compile
rake repackage
gem install pkg/nmatrix-*.gem

>> NMatrix.new([2, 3], [0, 1, 2, 3, 4, 5], :int64).pp
[0, 1, 2]
[3, 4, 5]
=> nil
>> m = N[ [2, 3, 4], [7, 8, 9] ]
=> #<NMatrix:0x007f8e121b6cf8shape:[2,3] dtype:int32
stype:dense>
>> m.pp
[2, 3, 4]
[7, 8, 9]
Depends on ATLAS/CBLAST
and written mostly in C and C++
https://github.com/SciRuby/nmatrix/wiki/Getting-started

Data Visualization
•R
•Gnuplot
•Google Charts API
•JFreeChart
•Scruffy
•Timetric
•Tioga
•RChart

Data Visualization
require 'rsruby'
cmd = %Q
(
pdf(file = "r_directly.pdf"))
boxplot(c(1,2,3,4),c(5,6,7,8))
dev.off()
)
def gnuplot(commands)
IO.popen("gnuplot", "w") { |io| io.puts commands }
end
commands = %Q(
set terminal svg
set output "curves.svg"
plot [-10:10] sin(x), atan(x), cos(atan(x))
)
gnuplot(commands)
http://effectif.com/ruby/manor/data-visualisation-with-ruby
https://github.com/glejeune/Ruby-Graphviz/Saturday, September 14, 2013

Other tools
•BioRuby
#!/usr/bin/env ruby

require 'bio'

# create a DNA sequence object from a String
dna = Bio::Sequence::NA.new("atcggtcggctta")

# create a RNA sequence object from a String
rna = Bio::Sequence::NA.new("auugccuacauaggc")

# create a Protein sequence from a String
aa = Bio::Sequence::AA.new("AGFAVENDSA")

# you can check if the sequence contains illegal characters
# that is not an accepted IUB character for that symbol
# (should prepare a Bio::Sequence::AA#illegal_symbols method also)
puts dna.illegal_bases

# translate and concatenate a DNA sequence to Protein sequence
newseq = aa + dna.translate
puts newseq # => "AGFAVENDSAIGRL"
http://bioruby.org/

Other tools
•RubyDoop (uses JRuby)
module
WordCount

class
Reducer

def
reduce(key,
values,
context)

sum
=
0

values.each
{
|value|
sum
+=
value.get
}

context.write(key,
Hadoop::Io::IntWritable.new(sum))

end

end
end
https://github.com/iconara/rubydoop
module
WordCount

class
Mapper

def
map(key,
value,
context)

value.to_s.split.each
do
|word|

word.gsub!(/W/,
'')

word.downcase!

unless
word.empty?

context.write(Hadoop::Io::Text.new(word),
Hadoop::Io::IntWritable.new(1))

end

end

end

end
end

Coming back to the
world of recommenders
The world is an over-crowded place

Coming back to the
world of recommenders!"#$%&'()$*+$,-$&.#'/0'&%)#)$1(,0#

Recommendation Systems
Systems designed to recommend to me something I may like

Recommendation Systems!"#$%&"'$"'(')*#*+,)
-+*#)+. -#/') 0#)1#
2' 23&4"+')1 5,6 7),*%'"&863
!
Graph Representation

And how does it work ?

What the recommenders realy do ?
1. Predict how much you may like a certain
product o service
2. It suggests a list of N items ordered by the level of
your interests.
3. It suggests a N list o f users to a product/
service
4. It explains to you why those items were
recommended.
5. It adjusts the prediction and recommendations
based on your feedback and from anothers.

Content Based Filtering
Gone with
the Wind
Die Hard
Similar
Armagedon
Toy
Store
Marcel
likes
recommends
Items
Users

Problems with Content
Recommenders
1. Restrict Data Analysis
3. Portfolio Effect
- Items and users mal-formed. Even worst in audio and images
- An person that does not have experience with Sushi does not get
the recommendation of the best sushi in town.
- Just because I saw 1 movie of Xuxa when I was child, it must have
to recommend all movies of her (só para baixinhos!)
2. Specialized Data

Collaborative Filtering
Gone with
the wind
Thor
Similar
Armagedon
Toy
Store
Marcel
like
recommend
Items
Rafael Amanda Users

Problems with Collaborative Filtering
1. Scalability
2. Sparse Data
3. Cold Start
4. Popularity
- Amazon with 5M users, 50K items, 1.4B ratings
- New users and items with no records
- I only rated one book at Amazon!
- The person who reads ‘Harry Potter’ also reads ‘Kama Sutra’
5. Hacking
- Everyone reads Harry Potter!

How does it show ?
Highlights More about this artist...
Listen to the similar songs
Someone similar to you also liked this...
Since you listened this, you may like this one...
Those items come together...
The most popular of your group...
New Releases

Recommendable
Quickly add a recommender engine for Likes and
Dislikes to your Ruby app
http://davidcel.is/recommendable/

Recommendable

Recommendable

gem
'recommendable'
Add to your GemFile:

Recommendable
require 'redis'
Recommendable.configure do |config|
# Recommendable's connection to Redis
config.redis = Redis.new(:host => 'localhost', :port => 6379, :db => 0)
# A prefix for all keys Recommendable uses
config.redis_namespace = :recommendable
# Whether or not to automatically enqueue users to have their
recommendations
# refreshed after they like/dislike an item
config.auto_enqueue = true
# The name of the queue that background jobs will be placed in
config.queue_name = :recommendable
# The number of nearest neighbors (k-NN) to check when updating
# recommendations for a user. Set to `nil` if you want to check all
# other users as opposed to a subset of the nearest ones.
config.nearest_neighbors = nil
end
Create a configuration initializer:

Recommendable
In your ONE model that will be receiving the
recommendations:
class User
recommends :movies, :books, :minerals,
:other_things
# ...
end

Recommendable
>> current_user.liked_movies.limit(10)
>> current_user.bookmarked_books.where(:author => "Cormac McCarthy")
>> current_user.disliked_movies.joins(:cast_members).where('cast_members.name = Kim Kardashian')
You can chain your queries

Recommendable
>> current_user.hidden_minerals.order('density DESC')
>> current_user.recommended_movies.where('year < 2010')
>> book.liked_by.order('age DESC').limit(20)
>> movie.disliked_by.where('age > 18')
You can chain your queries

Recommendable
You can also like your recommendable objects
>> user.like(movie)
=> true
>> user.likes?(movie)
=> true
>> user.rated?(movie)
=> true # also true if user.dislikes?(movie)
>> user.liked_movies
=> [#<Movie id: 23, name: "2001: A Space Odyssey">]
>> user.liked_movie_ids
=> ["23"]
>> user.like(book)
=> true
>> user.likes
=> [#<Movie id: 23, name: "2001: A Space Odyssey">, #<Book id: 42, title: "100 Years of Solitude">]
>> user.likes_count
=> 2
>> user.liked_movies_count
=> 1
>> user.likes_in_common_with(friend)
=> [#<Movie id: 23, name: "2001: A Space Odyssey">, #<Book id: 42, title: "100 Years of Solitude">]
>> user.liked_movies_in_common_with(friend)
=> [#<Movie id: 23, name: "2001: A Space Odyssey">]
>> movie.liked_by_count
=> 2
>> movie.liked_by
=> [#<User username: 'davidbowman'>, #<User username: 'frankpoole'>]

Recommendable
Obviously, You can also DISLIKE your recommendable
objects
>> user.dislike(movie)
>> user.dislikes?(movie)
>> user.disliked_movies
>> user.disliked_movie_ids
>> user.dislikes
>> user.dislikes_count
>> user.disliked_movies_count
>> user.dislikes_in_common_with(friend)
>> user.disliked_movies_in_common_with(friend)
>> movie.disliked_by_count
>> movie.disliked_by

Recommendable
Recommendations
>> friend.like(Movie.where(:name => "2001: A Space Odyssey").first)
>> friend.like(Book.where(:title => "A Clockwork Orange").first)
>> friend.like(Book.where(:title => "Brave New World").first)
>> friend.like(Book.where(:title => "One Flew Over the Cuckoo's Next").first)
>> user.like(Book.where(:title => "A Clockwork Orange").first)
=> [#<User username: "frankpoole">, #<User username: "davidbowman">, ...]
>> user.recommended_books # Defaults to 10 recommendations
=> [#<Book title: "Brave New World">, #<Book title: "One Flew Over the Cuckoo's
Nest">]
>> user.similar_raters # Defaults to 10 similar users
=> [#<
>> user.recommended_movies(10, 30) # 10 Recommendations, offset by 30 (i.e. page
4)
=> [#<Movie name: "A Clockwork Orange">, #<Movie name: "Chinatown">, ...]
>> user.similar_raters(25, 50) # 25 similar users, offset by 50 (i.e. page 3)
=> [#<User username: "frankpoole">, #<User username: "davidbowman">, ...]

Recommendable
Jaccard Similarity
Marcel likes A, B, C and dislikes D
Amanda likes A, B and dislikes C
Guilherme likes C, D and dislikes A
Flavio likes B, C, E and dislikes D
J(Marcel, Amanda) =
([A,B].size + [].size - [C].size - [].size) / [A,B,C,D].size
J(Marcel, Amanda) =
2 + 0 - 1 - 0 / 4 = 1/4 = 0.25

Recommendable
Jaccard Similarity
J(Marcel, Guilherme) =
([C].size + [].size - [A].size - [D].size) / [A,B,C,D].size
J(Marcel, Guilherme) =
1 + 0 - 1 - 1 / 4 = 1/4 = - 0.25

Recommendable
Jaccard Similarity
J(Marcel, Flavio) =
([B,C].size + [D].size - [].size - [].size) / [A,B,C,D, E].size
J(Marcel, Flavio) =
2 + 0 - 0 - 0 = 2/5 = 0.4

Recommendable
Jaccard Similarity
MostSimilar(Marcel) = [ (Flavio, 0.4) , (Amanda, 0.25) , (Guilherme, -0.25)]

Recommendable
Recommendations
>> Movie.top
=> #<Movie name: "2001: A Space Odyssey">
>> Movie.top(3)
=> [#<Movie name: "2001: A Space Odyssey">, #<Movie name: "A Clockwork Orange">,
#<Movie name: "The Shining">]
The best of your recommendable models
Wilson score confidence - Reddit Algorithm

Recommendable
Callbacks
class User < ActiveRecord::Base
has_one :feed
recommends :movies
after_like :update_feed
def update_feed(obj)
feed.update "liked #{obj.name}"
end
end
apotonick/hooks to implement callbacks for liking,
disliking, etc

Recommendable
Recommendable::Helpers::Calculations.update_similarities_for(user.id)
Recommendable::Helpers::Calculations.update_recommendations_for(user.id)
Manual recommendations

redis makes the magic!
Manual recommendations

Recommendable
module
Recommendable

module
Workers

class
Resque

include
::Resque::Plugins::UniqueJob
if
defined?(::Resque::Plugins::UniqueJob)

@queue
=
:recommendable

def
self.perform(user_id)

Recommendable::Helpers::Calculations.update_similarities_for(user_id)

Recommendable::Helpers::Calculations.update_recommendations_for(user_id)

end

end

end
end
Recommendations over Queueing System
Put the workers to do the job! (SideKiq, Resque, DelayedJob)

Recommended Books
SatnamAlag, Collective Intelligence in
Action, Manning Publications, 2009
Toby Segaran, Programming Collective
Intelligence, O'Reilly, 2007

Recommended Books
Exploring everyday things
with R and Ruby, Sau Chang,
O’Reilly, 2012

Recommended Course
https://www.coursera.org/course/recsys

Ruby developers, It does
exist
Web

Recommender Systems with Ruby (adding machine learning, statistics, etc)

More Related Content

Viewers also liked (20)

Similar to Recommender Systems with Ruby (adding machine learning, statistics, etc) (20)

More from Marcel Caraciolo (20)

Recently uploaded (20)

Recommender Systems with Ruby (adding machine learning, statistics, etc)