A Shilling Group Detection Framework Based on Deep Learning Techniques

Algorithm 1

Extract user features.

	Input: user-item rating matrix R
	Output: user feature matrix H
	Begin:
(1)	Randomly initialize weight matrices W and W′
(2)	repeat
(3)	for each do
(4)
(5)
(6)	end for
(7)	calculate the loss according to equations (11)–(15)
(8)	perform back-propagation to update the weight matrices
(9)	until the loss converges
(10)	for each do
(11)
(12)	end for
(13)	utilize all user vectors to construct user feature matrix H
(14)	return H
	End