Collaborative Filtering Annex to Make Predictions for an Unseen User

prediction_for_new_user.py

(...)
    print('Making predictions for an unseen user...')
    items_to_rate = [trainset.to_raw_iid(i) for i in trainset.all_items()]

    # We will consider this user to be ID 0 since that is not a valid ID for our real users
    user_id = 0

    # Make predictions
    predictions = []
    for iid in items_to_rate:
        prediction = algo.predict(user_id, iid)

        predictions.append([int(prediction.iid), prediction.est])

    # Sort the prediction from best match to worst match
    predictions.sort(key=lambda x: x[1], reverse=True)

    # Save the results to S3
    bucket.put_object(Body=str(predictions), Key='predictions/' + str(user_id))
(...)