run_utils.py

import numpy as np


def ndcg(top_k_dict, k_options, test, user_column, item_column):
    discounted_gain_per_k = np.array([1 / np.log2(i + 1) for i in range(1, max(k_options) + 1)])
    ideal_discounted_gain_per_k = [discounted_gain_per_k[:ind + 1].sum() for ind, k in enumerate(discounted_gain_per_k)]
    ndcg_per_user_per_k = {}
    for k in k_options:
        ndcg_per_user_per_k[k] = []
    for user, predictions in top_k_dict.items():
        positive_test_interactions = test[item_column][test[user_column] == user].values
        hits = np.in1d(predictions[:max(k_options)], positive_test_interactions)
        user_dcg = np.where(hits, discounted_gain_per_k[:len(hits)], 0)
        for k in k_options:
            user_ndcg = user_dcg[:k].sum() / ideal_discounted_gain_per_k[k - 1]
            ndcg_per_user_per_k[k].append(user_ndcg)
    return ndcg_per_user_per_k


def hr(top_k_dict, k_options, test, user_column, item_column):
    hr_per_user_per_k = {}
    for k in k_options:
        hr_per_user_per_k[k] = []
    for user, predictions in top_k_dict.items():
        positive_test_interactions = test[item_column][test[user_column] == user].values
        hits = np.in1d(predictions[:max(k_options)], positive_test_interactions)
        for k in k_options:
            user_hr = hits[:k].sum()
            user_hr = 1 if user_hr > 0 else 0
            hr_per_user_per_k[k].append(user_hr)
    return hr_per_user_per_k


def recall(top_k_dict, k_options, test, user_column, item_column):
    recall_per_user_per_k = {}
    for k in k_options:
        recall_per_user_per_k[k] = []
    for user, predictions in top_k_dict.items():
        positive_test_interactions = test[item_column][test[user_column] == user].values
        hits = np.in1d(predictions[:max(k_options)], positive_test_interactions)
        for k in k_options:
            if user == 8:
                pass
            user_recall = hits[:k].sum() / min(len(positive_test_interactions), k)
            recall_per_user_per_k[k].append(user_recall)
    return recall_per_user_per_k