[BUG] relevancy method and get_top_k with threshold discussion

[miguelgfierro]

Description

@anargyri, @loomlike and I were having a discussion about how to optimize surprise prediction, and we saw that in the function merge_ranking_true_pred there is this code:

    if relevancy_method == "top_k":
        top_k = k
    elif relevancy_method == "by_threshold":
        top_k = threshold
    else:
        raise NotImplementedError("Invalid relevancy_method")
    df_hit = get_top_k_items(
        dataframe=rating_pred_common,
        col_user=col_user,
        col_rating=col_prediction,
        k=top_k,
    )

so the input to get_top_k_items is a number, but not the relevancy method. In that function we have:

    top_k_items = (
        dataframe.groupby(col_user, as_index=False)
        .apply(lambda x: x.nlargest(k, col_rating))
        .reset_index(drop=True)
    )

@yueguoguo are we using the threshold somewhere?

In which platform does it happen?

How do we replicate the issue?

Expected behavior (i.e. solution)

Other Comments