Overview : Tensorflow/Keras implementation of sequential recommendation system using convolutionally enhanced item embeddings and triplet loss.

For more information : https://arxiv.org/pdf/1809.07426.pdf

MovieLens 1M Dataset : https://grouplens.org/datasets/movielens/1m/
for simplicity dataset located in my github repo : https://raw.githubusercontent.com/malinphy/datasets/main/ml_1M/ratings.dat

• data_prep.py : generation of negative samples and target values
• HelperFunctions.py : Data preparation for model training
• model.py : Caser model written with tensoflow/keras
• train.py : training file
• caser_model_weights.h5 : model weights
• eval.py : mean average precision MAP calculation
• requirements.txt : required packages and versions to run model = caser_prediction.py : prediction file for deployment purpose

download file

conda create --name caser_env
conda activate revenue_model

find the folder directory in caser_env

pip install -r requirements.txt 

run train.py file
for deployment purpose prediction file created seperately as caser_prediction.py

original study investigated up to next 3 items. However, this implementation designed for 1 item. Designing will be developed according to original model.

If you use this Caser in your paper, please cite the paper:

@inproceedings{tang2018caser,
  title={Personalized Top-N Sequential Recommendation via Convolutional Sequence Embedding},
  author={Tang, Jiaxi and Wang, Ke},
  booktitle={ACM International Conference on Web Search and Data Mining},
  year={2018}
}