Implementation of implicit feedback model - Githubissues

ADS2024-B2 / AgileProject

0 stars 0 forks source link

Implementation of implicit feedback model #42

Open cfuses opened 1 week ago

cfuses commented 1 week ago

In sprint 1 we succesfully implemented the explicit feedback recommender, now we will implement the implicit feedback model from spotlight, as it is also a model with no parameters to be saved. This is a good option for having a better model generating our app recommendations, while not having the resources to save model parameters in mlflow yet.

Testing:

[x] The notebook training the model is executed without errors.
[x] A new model is correctly saved and can be loaded back into a python script.

Related to User story #40

cfuses commented 2 days ago

New model seems to perform worse than explicit model: Implicit: Train RMSE 2.668, test RMSE 2.678 Explicit: Train RMSE 0.893, test RMSE 0.938

We should discuss whether we should invest more time fine tunning this new model, as research says it should work better than the explicit one, or directly go for the sequential model.