def test_best_avg_recency_weight_lesser_is_better():
df = pd.DataFrame.from_dict(
{
"model_group_id": ["1", "2", "3", "1", "2", "3", "1", "2", "3"],
"model_id": ["1", "2", "3", "4", "5", "6", "7", "8", "9"],
"train_end_time": [
"2011-01-01",
"2011-01-01",
"2011-01-01",
"2012-01-01",
"2012-01-01",
"2012-01-01",
"2013-01-01",
"2013-01-01",
"2013-01-01",
],
"metric": [
"false positives@",
"false positives@",
"false positives@",
"false positives@",
"false positives@",
"false positives@",
"false positives@",
"false positives@",
"false positives@",
],
"parameter": [
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
],
"raw_value": [20, 90, 40, 50, 50, 50, 80, 20, 50],
"dist_from_best_case": [70, 0, 50, 0, 0, 0, 0, 60, 30],
}
)
df["train_end_time"] = pd.to_datetime(df["train_end_time"])
assert best_avg_recency_weight(
df, "2013-01-01", "false positives@", "100_abs", 1.00, "linear"
) == ["3"]
assert best_avg_recency_weight(
df, "2013-01-01", "false positives@", "100_abs", 1.15, "linear"
) == ["3"]
assert best_avg_recency_weight(
df, "2013-01-01", "false positives@", "100_abs", 1.15, "linear", n=2
) == ["3", "1"]
assert best_avg_recency_weight(
df, "2013-01-01", "false positives@", "100_abs", 1.50, "linear"
) == ["3"]
assert best_avg_recency_weight(
df, "2013-01-01", "false positives@", "100_abs", 1.50, "linear", n=2
) == ["3", "2"]
def test_best_avg_recency_weight_greater_is_better():
df = pd.DataFrame.from_dict(
{
"model_group_id": ["1", "2", "3", "1", "2", "3", "1", "2", "3"],
"model_id": ["1", "2", "3", "4", "5", "6", "7", "8", "9"],
"train_end_time": [
"2011-01-01",
"2011-01-01",
"2011-01-01",
"2012-01-01",
"2012-01-01",
"2012-01-01",
"2013-01-01",
"2013-01-01",
"2013-01-01",
],
"metric": [
"precision@",
"precision@",
"precision@",
"precision@",
"precision@",
"precision@",
"precision@",
"precision@",
"precision@",
],
"parameter": [
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
"100_abs",
],
"raw_value": [0.8, 0.2, 0.4, 0.5, 0.5, 0.5, 0.2, 0.7, 0.5],
"dist_from_best_case": [0.0, 0.4, 0.2, 0.0, 0.0, 0.0, 0.5, 0.0, 0.2],
}
)
df["train_end_time"] = pd.to_datetime(df["train_end_time"])
assert best_avg_recency_weight(
df, "2013-01-01", "precision@", "100_abs", 1.00, "linear"
) == ["1"]
assert best_avg_recency_weight(
df, "2013-01-01", "precision@", "100_abs", 1.00, "linear", n=2
) == ["1", "2"]
assert best_avg_recency_weight(
df, "2013-01-01", "precision@", "100_abs", 1.15, "linear"
) == ["1"]
assert best_avg_recency_weight(
df, "2013-01-01", "precision@", "100_abs", 1.50, "linear"
) == ["2"]
assert best_avg_recency_weight(
df, "2013-01-01", "precision@", "100_abs", 1.50, "linear", n=2
) == ["2", "3"]
def test_best_avg_recency_weight_lesser_is_better():
df = pandas.DataFrame.from_dict({
'model_group_id': ['1', '2', '3', '1', '2', '3', '1', '2', '3'],
'model_id': ['1', '2', '3', '4', '5', '6', '7', '8', '9'],
'train_end_time': [
'2011-01-01', '2011-01-01', '2011-01-01', '2012-01-01',
'2012-01-01', '2012-01-01', '2013-01-01', '2013-01-01',
'2013-01-01'
],
'metric': [
'false positives@', 'false positives@', 'false positives@',
'false positives@', 'false positives@', 'false positives@',
'false positives@', 'false positives@', 'false positives@'
],
'parameter': [
'100_abs', '100_abs', '100_abs', '100_abs', '100_abs', '100_abs',
'100_abs', '100_abs', '100_abs'
],
'raw_value': [20, 90, 40, 50, 50, 50, 80, 20, 50],
'dist_from_best_case': [70, 0, 50, 0, 0, 0, 0, 60, 30],
})
df['train_end_time'] = pandas.to_datetime(df['train_end_time'])
assert best_avg_recency_weight(df, '2013-01-01', 'false positives@',
'100_abs', 1.00, 'linear') == ['3']
assert best_avg_recency_weight(df, '2013-01-01', 'false positives@',
'100_abs', 1.15, 'linear') == ['3']
assert best_avg_recency_weight(df,
'2013-01-01',
'false positives@',
'100_abs',
1.15,
'linear',
n=2) == ['3', '1']
assert best_avg_recency_weight(df, '2013-01-01', 'false positives@',
'100_abs', 1.50, 'linear') == ['3']
assert best_avg_recency_weight(df,
'2013-01-01',
'false positives@',
'100_abs',
1.50,
'linear',
n=2) == ['3', '2']
def test_best_avg_recency_weight_greater_is_better():
df = pandas.DataFrame.from_dict({
'model_group_id': ['1', '2', '3', '1', '2', '3', '1', '2', '3'],
'model_id': ['1', '2', '3', '4', '5', '6', '7', '8', '9'],
'train_end_time': [
'2011-01-01', '2011-01-01', '2011-01-01', '2012-01-01',
'2012-01-01', '2012-01-01', '2013-01-01', '2013-01-01',
'2013-01-01'
],
'metric': [
'precision@', 'precision@', 'precision@', 'precision@',
'precision@', 'precision@', 'precision@', 'precision@',
'precision@'
],
'parameter': [
'100_abs', '100_abs', '100_abs', '100_abs', '100_abs', '100_abs',
'100_abs', '100_abs', '100_abs'
],
'raw_value': [0.8, 0.2, 0.4, 0.5, 0.5, 0.5, 0.2, 0.7, 0.5],
'dist_from_best_case': [0.0, 0.4, 0.2, 0.0, 0.0, 0.0, 0.5, 0.0, 0.2],
})
df['train_end_time'] = pandas.to_datetime(df['train_end_time'])
assert best_avg_recency_weight(df, '2013-01-01', 'precision@', '100_abs',
1.00, 'linear') == ['1']
assert best_avg_recency_weight(df,
'2013-01-01',
'precision@',
'100_abs',
1.00,
'linear',
n=2) == ['1', '2']
assert best_avg_recency_weight(df, '2013-01-01', 'precision@', '100_abs',
1.15, 'linear') == ['1']
assert best_avg_recency_weight(df, '2013-01-01', 'precision@', '100_abs',
1.50, 'linear') == ['2']
assert best_avg_recency_weight(df,
'2013-01-01',
'precision@',
'100_abs',
1.50,
'linear',
n=2) == ['2', '3']