In this repository an example code is presented on how to put a LightFM
hybrid recommendation
system into production.
The MovieLens
dataset is used to train a LightFM
model. Then it is illustrated how to quickly
incorporate new interactions to the model, without completely retraining it, using the
fit_partial
function. Finally, it is showed how to prepare the model to anticipate new
users / items, by padding the training data with some dummy users / items.
Create a Python 3 environment and install the dependencies from requirements.txt
.
Open the recommendation_to_production.ipynb
notebook.