Online Doctor Recommendation with Convolutional Neural Network and Sparse Inputs

Algorithm 1

PMF-CNN algorithm.

	Input: Interaction matrix between patients’ reviews and doctors’ categories
	Output: The predictive scoring matrix of patients’ reviews and doctors’ categories
(1)	fordo
(2)	Select a review matrix randomly from the reviews for training;
(3)	the training batch size is , and the size of each batch is ;
(4)	calculate the loss in the training process;
(5)	if or is small enough then
(6)	Use (5), (6), and (9) to calculate the predicted score value of each instance;
(7)	Optimizing the parameters of the model by minimizing (3);
(8)	end if
(9)	end for