context_recommendation

Contextual Recommendation Implementation for Research Purposes.

Environment

python: python2/python3 tensorflow: 1.6.0 numpy: 1.15.3 cPickle/_pickle: 1.71

1. Dataset

Yelp Dataset(The experiment in paper is conducted on Version 7). Latest dataset (version 9) can be downloaded from: https://www.kaggle.com/yelp-dataset/yelp-dataset. (There are some minor changes in different versions of dataset, e.g. 'neighbourhood' features, etc.)

Unzip the archive file and you can see: yelp_academic_dataset_review.json, yelp_academic_dataset_business.jsonm, yelp_academic_dataset_user.json

Yelp Dataset(version 7) is no longer on the kaggle website. You can download dataset (version 7) of the paper's experiment and pretrain/train/test .pkl pickle files from the cloud link as below: Link: https://pan.baidu.com/s/1EUm05wn88bDru-LTxuCETw, Password:f1ej

# Generate Pretraining/Train/Test Examples:
python contextual_dataset_yelp.py

# output multiple files 
# Pretrain: ../data/yelp/yelp-dataset/yelp_pretrain_dataset.pkl
# train:    ../data/yelp/yelp-dataset/train/yelp_train_examples_*.pkl   * is the 1-10
# test:     ../data/yelp/yelp-dataset/test/yelp_test_examples_*.pkl     * is the 1-2

You can also download these data files(version 7) since generating dataset is time-consuming.
Link: https://pan.baidu.com/s/1EUm05wn88bDru-LTxuCETw
Password:f1ej

python contextual_dataset_yelp_old.py

2. Pretraining using MACDAE model on Yelp Dataset

# Pretrain the recommendation model using multi-heads number as 4
python contextual_macdae_yelp.py 4

3. Training Recommendation model

# BASE: Wide&Deep
python recommd_context_pretrain_yelp.py base train

# BASE+MACDAE: Wide&Deep + Pretrain the recommendation model using multi-heads number as 4
python recommd_context_pretrain_yelp.py macdae train

4. Evaluating Recommendation model

## BASE(Wide & Deep), input x dimension (339)
# Example Test Run Result: NDCG@5 is 0.3827, NDCG@10 is 0.4348

python recommd_context_pretrain_yelp.py base test

## BASE(Wide & Deep) + MACDAE， input x dimension (403)
# Example Test Run Result: NDCG@5 is 0.4382, NDCG@10 is 0.4854

python recommd_context_pretrain_yelp.py macdae test

Related

References

http://www.deepnlp.org/blog/
http://www.deepnlp.org/equation/
http://www.deepnlp.org/search/
http://www.deepnlp.org/workspace/ai_courses/
http://www.deepnlp.org/workspace/aigc_chart/
http://www.deepnlp.org/workspace/ai_writer/
http://www.deepnlp.org/workspace/detail/
AI IMAGE GENERATOR
AI Search Engine
AI Chatbot Assistant
AI for ELDERLY
AI for KIDS
AI in LAW
AI in FINANCE
AI in HEALTHCARE
AI in BUSINESS
AI in EDUCATION
AI in PRODUCTIVITY TOOL
AI in POLITICS
AI in ENTERTAINMENT
AI in NEWS
AI in ART AND SPORTS
AI in LIFESTYLE
AI in PAYMENT
AI in SOCIAL
AI in AGRICULTURE
AI in SCIENCE
AI in TECHNOLOGY
AI in TRAVEL
AI in TRANSPORTATION
AI in CAR
AI in CHARITY
AI in PUBLIC SERVICE
AI in HOUSING
AI in COMMUNICATION
AI in FOOD

Related Blog

http://www.deepnlp.org/blog/
http://www.deepnlp.org/equation/
http://www.deepnlp.org/search/
http://www.deepnlp.org/workspace/ai_courses/
http://www.deepnlp.org/workspace/aigc_chart/
http://www.deepnlp.org/workspace/ai_writer/
http://www.deepnlp.org/workspace/detail/
Statistics Equation Formula
Machine Learning Equation Formula
AI Courses for Kids
AI in Fashion: Tell IWC Schaffhausen Watches Real or Fake
AI in Fashion: Tell Fendi bags real or fake
AI in Fashion: Tell Coach bags real or fake
AI in Fashion: Tell Prada bags real or fake
AI in Fashion: Tell Gucci bags real or fake
AI in Fashion: Tell Dior bags real or fake
AI in Fashion: Tell Hermes bags real or fake
AI in Fashion: Tell Chanel bags real or fake
AI in Fashion: Tell Louis Vuitton bags real or fake
AI in Fashion: Tell Omega Watches real or fake
AI in Fashion: Tell Rolex Watches real or fake