A Multidimensional Nonnegative Matrix Factorization Model for Retweeting Behavior Prediction

Algorithm 1

Multidimensional nonnegative matrix factorization model for retweeting behavior prediction.

Input: retweeting matrices ; user-based features matrix ; content-based features matrix ;

parameters , , , ,

Output: predicted retweeting matrix

The description of the algorithm:

(1) for each do

(2)  for each do

(3)   if or then

(4)    calculate with (10)

(5)   end if

(6)  end for

(7) end for

(8) Initialize with and set elements which are negative in to 0

(9) Initialize with and set elements which are negative in to 0

(10) Repeat

(11)  update and with (20)

(12) Until in (14) reaches to convergence

(13) Repeat

(14)  update and with (22)

(15) Until in (21) reaches to convergence

(16) Calculate error rate of :

   % denotes the number of instances which are wrongly classified with

(17) Calculate error rate of :

   % denotes the number of instances which are wrongly classified with

(18) Return