CMSC12300_Project Yelp Recommendation

Group members:
Ying Sun
Jiaxu Han
Fulin Guo
Ellen Hsieh

How to run:

In terminal:

Clustering Algorithm:

python3 cluster_model_1.py -r dataproc --num-core-instances 6 cluster_data.csv

python3 cluster_model_2.py -r dataproc --num-core-instances 6 cluster_data.csv

python3 cluster_model_3.py -r dataproc --num-core-instances 6 cluster_data.csv

python3 cluster_model_4.py -r dataproc --num-core-instances 6 cluster_data.csv

User-to-user Algorithm:

python3 user_to_user_model_1.py -r dataproc --num-core-instances 6 user_to_user_data.csv

python3 user_to_user_model_2.py -r dataproc --num-core-instances 6 user_to_user_data.csv

python3 user_to_user_model_3.py -r dataproc --num-core-instances 6 user_to_user_data.csv

python3 user_to_user_model_4.py -r dataproc --num-core-instances 6 user_to_user_data.csv

Friend recommendation:

python3 friend_recommendation.py -r dataproc --num-core-instances 6 friend_recommendation_data.csv

Note.

cluster_data.csv can be found in Data folder

user_to_user_data.csv can be downloaded from a public bucket gs://hanjiaxu/user_to_user.csv (see Data folder readme)

friend_recommendation_data.csv can be found in Data folder

Other .py files can be found in Code folder

Parameters

cluster_model_1:

Frequency threshold: 1 Rating threshold: 2 Times threshold: 1

cluster_model_2:

Frequency threshold: 1 Rating threshold: 3 Times threshold: 1

cluster_model_3:

Frequency threshold: 1; Rating threshold: 3; Times threshold: 2

cluster_model_4:

Frequency threshold: 1; Rating threshold: 4; Times threshold: 1

user_to_user_model_1:

Weight: 0.5; Top 5 similar users

user_to_user_model_2:

Weight: 0.7； Top 5 similar users

user_to_user_model_3:

Weight: 0.5； Top 3 similar users

user_to_user_model_4:

Weight: 0.7； Top 3 similar users

Result

Please refer to the final report for specific meanings about the parameters