cathy27 / dsin

Code for the IJCAI'19 paper "Deep Session Interest Network for Click-Through Rate Prediction"

Deep Session Interest Network for Click-Through Rate Prediction

Experiment code on Advertising Dataset of paper Deep Session Interest Network for Click-Through Rate Prediction(https://arxiv.org/abs/1905.06482)

Yufei Feng , Fuyu Lv, Weichen Shen and Menghan Wang and Fei Sun and Yu Zhu and Keping Yang.

In Proceedings of 28th International Joint Conference on Artificial Intelligence (IJCAI 2019)

---

Operating environment

• python==3.6
• tensorflow-gpu==1.4.0
• deepctr==0.4.1
• numpy==1.15.1
• pandas==0.22.0
• scikit-learn==0.19.2
• tqdm==4.19.5

---

Download dataset and preprocess

Download dataset

1. Download Dataset Ad Display/Click Data on Taobao.com
2. Extract the files into the raw_data directory

Data preprocessing

1. run 0_gen_sampled_data.py, sample the data by user
2. run 1_gen_sessions.py, generate historical session sequence for each user

Training and Evaluation

Train DIN model

1. run 2_gen_din_input.py,generate input data
2. run train_din.py

Train DIEN model

1. run 2_gen_dien_input.py,generate input data(It may take a long time to sample negative samples.)
2. run train_dien.py

Train DSIN model

1. run 2_gen_dsin_input.py,generate input data
2. run train_dsin.py

   The loss of DSIN with bias_encoding=True may be NaN sometimes on Advertising Dataset and it remains a confusing problem since it never occurs in the production environment.We will work on it and also appreciate your help.

License

This project is licensed under the terms of the Apache-2 license. See LICENSE for additional details.