matrix_factorization.rb

class MF
	attr_accessor :base_file, :tag_file, :train_users, :test_users, :genders, :ratings, :items, :users, :r_matrix

	def initialize()
		@base_file = File.read("u.data")
		@tag_file = File.read("u.user")

		@train_users = Array.new
		@test_users = Array.new
		@users = Array.new
		@genders = Hash.new
		@items = Array.new
		@ratings = Hash.new

		@k = 1
	end


	def sigmoid(x)
		return 1.0 / (1 + Math.exp(-x))
	end

	def read_tag_file()
		splitted = Array.new
		@tag_file.each_line.with_index do |line, i|
			splitted = line.split("|")
			user = splitted[0].to_i
			gender = splitted[2]
			@genders[user] = gender
		end
	end

	def read_base_file()
		@base_file.each_line.with_index do |line, i|
			splitted = line.split(" ")
			user = splitted[0].to_i
			item = splitted[1].to_i
			rating = splitted[2].to_i
			time_stamp = splitted[3].to_i

			if(time_stamp <= 880845177)
				if(not train_users.include? user)
					@train_users << user
					@ratings[user] = Hash.new
				end

				if(not items.include? item)
					@items << item
				end
				@ratings[user][item] = rating
				
			elsif(time_stamp <= 884673930)
				if(not test_users.include? user)
					@test_users << user
					@ratings[user] = Hash.new
				end

				if(not items.include? item)
					@items << item
				end
				@ratings[user][item] = rating
			end
		end

		@test_users -= @train_users
		@users = @train_users + @test_users

	end


	def edit_rating()
		@r_matrix = Array.new(@users.size) {Array.new(@items.size)}
		@users.each.with_index do |user, u_i|
			@items.each.with_index do |item, i_i|
				if(@ratings[user][item].nil?)
					@r_matrix[u_i][i_i] = 0
				# elsif(@ratings[user][item] > 3)
					# @r_matrix[u_i][i_i] = 1
				# elsif(@ratings[user][item] > 3)
					# @r_matrix[u_i][i_i] = 2
				else
					@r_matrix[u_i][i_i] = 1
				end
			end
		end
	end

	def learn()

		step_size = 0.001
		theta = 0.0001
		steps = 150
		mu = 0.5
		
		alpha = Array.new(@users.size, 0)
		betha = Array.new(@items.size, 0)

		@p_matrix = Array.new(@users.size) {Array.new(@k, 0)}
		q_matrix = Array.new(@k) {Array.new(@items.size, 0)}

		alpha.each.with_index do |user, i|
			alpha[i] = rand
		end

		betha.each.with_index do |user, i|
			betha[i] = rand
		end

		@p_matrix.each.with_index do |array, i|
			array.each.with_index do |b, j|
				@p_matrix[i][j] = rand
			end
		end


		q_matrix.each.with_index do |array, i|
			array.each.with_index do |b, j|
				q_matrix[i][j] = rand
			end
		end

		p_matrix_old = Array.new(@k, 0)

		
		steps.times do |iteration|
			if(iteration % 10 == 0)
				puts iteration
			end
			@users.each.with_index do |user, i|
				@items.each.with_index do |item, j|
					if(@r_matrix[i][j] != 0)
						sigma = 0.0
						@k.times do |counter|
							sigma += @p_matrix[i][counter] * q_matrix[counter][j]
						end

						r_prime = sigmoid(mu + alpha[i] + betha[j] + sigma)
						diff = (@r_matrix[i][j] - r_prime)

						@k.times do |counter|
							p_matrix_old[counter] = @p_matrix[i][counter]
						end

						@k.times do |counter|
							@p_matrix[i][counter] += step_size * (diff * q_matrix[counter][j] - theta * @p_matrix[i][counter])
							q_matrix[counter][j] += step_size * (diff * p_matrix_old[counter] - theta * q_matrix[counter][j])
						end

						alpha[i] += step_size * diff
						betha[j] += step_size * diff
						mu += step_size * diff
					end
				end
			end
		end

		def write_in_file()
			mf_train = File.open("mf_train#{@k}.csv", "w")
			mf_test = File.open("mf_test#{@k}.csv", "w")

			feature_line =  ""
			@k.times do |counter|
				feature_line << counter.to_s
				feature_line << ","
			end

			feature_line << "label"

			mf_train.puts(feature_line)
			mf_test.puts(feature_line)

			@users.each.with_index do |user, i|
				new_line = ""
				@k.times do |counter|
					new_line << @p_matrix[i][counter].to_s
					new_line << ","
				end
				new_line << genders[users[i]].to_s

				if(@train_users.include? users[i])
					mf_train.puts(new_line)	
				elsif(@test_users.include? users[i])
					mf_test.puts(new_line)
				else
					puts "error"
				end
			end
		end
	end
end

mf = MF.new
mf.read_tag_file()
mf.read_base_file()
mf.edit_rating()
mf.learn()
mf.write_in_file()