Research Article

A Time-Aware CNN-Based Personalized Recommender System

Algorithm 1

Time-aware CNN-based personalized recommendation algorithm.
Input: users.dat, items.dat, ratings.dat with timestamp Rt, target user u, user u’s time context t
Output: recommended item list for u RL
Step 1: process data and save processed data to preprocess.p;
Step 2: open preprocess.p and set parameters;
Step 3: construct NN and generate users’ features and items’ features;
Step 4: construct graph to calculate prediction rating by user similarity calculation, update parameter settings according to MSE;
Step 5: randomly split dataset into training set and test set, and then train NN;
Step 6: save trained model and parameters;
Step 7: load saved model to recommend for target user u according to t;
Step 8: generate recommendation list RL from time-aware CNN-based personalized recommendation algorithm;
Step 9: return RL.