def test_calculating_ndcg_test_index():
case1 = [4, 3, 2, 0, 1]
dcg_score_1 = ndcg_at_k(case1, 3)
dcg_score_2 = ndcg_at_k(case1, 4)
dcg_score_3 = ndcg_at_k(case1, 5)
assert [dcg_score_1, dcg_score_3, dcg_score_2] == sorted(
[dcg_score_2, dcg_score_1, dcg_score_3], reverse=True)
def test_calculating_ndcg_non_ideal_case():
case1 = [4, 3, 2, 0, 1]
case2 = [4, 2, 3, 0, 1]
case3 = [0, 1, 2, 3, 4]
dcg_score_1 = ndcg_at_k(case1, 5)
dcg_score_2 = ndcg_at_k(case2, 5)
dcg_score_3 = ndcg_at_k(case3, 5)
assert [dcg_score_1, dcg_score_2, dcg_score_3] == sorted(
[dcg_score_2, dcg_score_1, dcg_score_3], reverse=True)
def calculate_relevancy(ground_truth, search_results):
if not (ground_truth and search_results):
logging.warn("Parameters are empty")
return [(0.0, 0)]
relevancy_map = {x: i + 1 for i, x in enumerate(reversed(ground_truth))}
logging.debug("Found relevancy {}".format(relevancy_map))
rel = [relevancy_map.get(result, 0) for result in search_results]
logging.debug("Found relevancy {}".format(rel))
starting_value = min(3, len(ground_truth))
ranges = [starting_value] + [i for i in [3, 5, 10, 15, 25, 50, 100] if i > starting_value]
scores = [(ndcg_at_k(rel, i), i) for i in ranges
if i > starting_value]
logging.debug("Found relevancy score {}".format(scores))
return scores
def test_calculating_ndcg_extra_k():
case1 = [4, 3, 2, 0]
dcg_score_1 = ndcg_at_k(case1, 5)
print(dcg_score_1)
def test_calculating_ndcg():
ground_truth = [4, 3, 2, 1, 0]
dcg_score_ideal = ndcg_at_k(ground_truth, 5)
assert dcg_score_ideal == 1.0