Movie recommender system based on Deep Factorization Machine, DeepFM [1] . It combines matrix factorization and deep neural network to build a hybrid recommendation system. The neural network captures the high-order features and the factorization method captures the low-order features.
Let the input be x , output of the factorization method be yMF(x) and the output of the neural network be yNN(x). Then the final prediction is given by,
where is the sigmoid activation function. For more details, refer to [1].
To train the network, run the following command in terminal
python train.py
It will train the network using default settings, show the training loss figure and save the trained model as 'model.pt'.
training mse: 0.01361, testing mse: 0.05956
[1]: Guo, Huifeng, Ruiming Tang, Yunming Ye, Zhenguo Li, and Xiuqiang He. "Deepfm: a factorization-machine based neural network for ctr prediction." arXiv preprint arXiv:1703.04247 (2017).