A PyTorch implementation of DeepFM for CTR prediction problem.

1. Download Criteo's Kaggle display advertising challenge dataset from here( if you have had it already, skip it ), and put it in ./data/raw/

2. Generate a preprocessed dataset.

    ./utils/dataPreprocess.py
   

3. Train a model and predict.

    ./main.py
   

• https://github.com/nzc/dnn_ctr.

• https://github.com/PaddlePaddle/models/tree/develop/deep_fm.

• DeepFM: A Factorization-Machine based Neural Network for CTR Prediction, Huifeng Guo, Ruiming Tang, Yunming Yey, Zhenguo Li, Xiuqiang He.