This is our implementation for the paper:

Yihao Zhang, Meng Yuan, Chu Zhao, Mian Chen and Xiaoyang Liu. Aggregating Knowledge-aware Graph Neural Network and Adaptive Relational Attention for Recommendation.

tensorflow = 1.12 numpy = 1.18

python preprocess.py 

python deal_data.py

python main.py

We provide three processed datasets: MovieLens 1 Million (ml-1m), LastFM, and BookCrossing.

kg.txt:

• knowledge graph file;

item_index2entity_id.txt:

• the mapping from item indices in the raw rating file to entity IDs in the KG;

user_artists.dat: raw rating file of Last.FM

ratings_final.txt:

• user item interaction file

train.txt:

• train file
• each Line is a training instance: userID\t itemID\t rating

test.txt:

• test file (positive instances)
• each Line is a testing instance: userID\t itemID\t rating

test_negative.txt:

• test file (negative instances)
• each line corresponds to the line of test.rating, containing 100 negative samples.
• each line is in the format: (userID,itemID)\t negativeItemID1\t negativeItemID2 ...