flurs.utils.metric

Ranking-based evaluation metrics for recommender systems

Functions

`auc`(truth, recommend)	Area under the ROC curve (AUC).
`average_precision`(truth, recommend)	Average Precision (AP).
`count_true_positive`(truth, recommend)	Count number of true positives from given sets of samples.
`mpr`(truth, recommend)	Mean Percentile Rank (MPR).
`ndcg`(truth, recommend[, k])	Normalized Discounted Cumulative Grain (NDCG).
`precision`(truth, recommend[, k])	Precision@k.
`recall`(truth, recommend[, k])	Recall@k.
`reciprocal_rank`(truth, recommend)	Reciprocal Rank (RR).

flurs.utils.metric.auc(truth, recommend)[source]

Area under the ROC curve (AUC).

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.

Returns

AUC.

Return type

float

flurs.utils.metric.average_precision(truth, recommend)[source]

Average Precision (AP).

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.

Returns

Average Precision.

Return type

float

flurs.utils.metric.count_true_positive(truth, recommend)[source]

Count number of true positives from given sets of samples.

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.

Returns

Number of true positives.

Return type

int

flurs.utils.metric.mpr(truth, recommend)[source]

Mean Percentile Rank (MPR).

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.

Returns

Mean Percentile Rank.

Return type

float

flurs.utils.metric.ndcg(truth, recommend, k=None)[source]

Normalized Discounted Cumulative Grain (NDCG).

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.
k (int or None, default=None) – Top-k items in recommend are considered to be recommended. Defaults to len(recommend).

Returns

NDCG@k.

Return type

float

flurs.utils.metric.precision(truth, recommend, k=None)[source]

Precision@k.

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.
k (int or None, default=None) – Top-k items in recommend are considered to be recommended. Defaults to len(recommend).

Returns

Precision@k.

Return type

float

flurs.utils.metric.recall(truth, recommend, k=None)[source]

Recall@k.

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.
k (int or None, default=None) – Top-k items in recommend are considered to be recommended. Defaults to len(recommend).

Returns

Recall@k.

Return type

float

flurs.utils.metric.reciprocal_rank(truth, recommend)[source]

Reciprocal Rank (RR).

Parameters

truth (numpy 1d array) – Set of truth samples.
recommend (numpy 1d array) – Ordered listed of recommended samples.

Returns

Reciprocal Rank.

Return type

float