C# Class MyMediaLite.Eval.Items

Evaluation class for item recommendation

Datei anzeigen Open project: zenogantner/MML-KDD

Public Methods

Method	Description
AUC ( int ranked_items, ICollection correct_items ) : double	Compute the area under the ROC curve (AUC) of a list of ranked items
AUC ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double	Compute the area under the ROC curve (AUC) of a list of ranked items
DisplayResults ( double>.Dictionary result ) : void	Display item prediction results
Evaluate ( IItemRecommender recommender, IPosOnlyFeedback test, IPosOnlyFeedback train, ICollection relevant_users, ICollection relevant_items ) : double>.Dictionary	Evaluation for rankings of items User-item combinations that appear in both sets are ignored for the test set, and thus in the evaluation. The evaluation measures are listed in the ItemPredictionMeasures property. Additionally, 'num_users' and 'num_items' report the number of users that were used to compute the results and the number of items that were taken into account. Literature: C. Manning, P. Raghavan, H. Schütze: Introduction to Information Retrieval, Cambridge University Press, 2008
EvaluateOnline ( IItemRecommender recommender, IPosOnlyFeedback test, IPosOnlyFeedback train, ICollection relevant_users, ICollection relevant_items ) : double>.Dictionary	Online evaluation for rankings of items
MAP ( int ranked_items, ICollection correct_items ) : double	Compute the mean average precision (MAP) of a list of ranked items
MAP ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double	Compute the mean average precision (MAP) of a list of ranked items
NDCG ( int ranked_items, ICollection correct_items ) : double	Compute the normalized discounted cumulative gain (NDCG) of a list of ranked items
NDCG ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double	Compute the normalized discounted cumulative gain (NDCG) of a list of ranked items
PrecisionAt ( int ranked_items, ICollection correct_items, ICollection ignore_items, int n ) : double	Compute the precision@N of a list of ranked items
PrecisionAt ( int ranked_items, ICollection correct_items, int n ) : double	Compute the precision@N of a list of ranked items

Private Methods

Method	Description
ComputeIDCG ( int n ) : double	Computes the ideal DCG given the number of positive items.

Method Details

AUC() public static method

Compute the area under the ROC curve (AUC) of a list of ranked items

public static AUC ( int ranked_items, ICollection correct_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
return	double

AUC() public static method

Compute the area under the ROC curve (AUC) of a list of ranked items

public static AUC ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
ignore_items	ICollection	a collection of item IDs which should be ignored for the evaluation
return	double

DisplayResults() public static method

Display item prediction results

public static DisplayResults ( double>.Dictionary result ) : void
result	double>.Dictionary	the result dictionary
return	void

Evaluate() public static method

Evaluation for rankings of items

User-item combinations that appear in both sets are ignored for the test set, and thus in the evaluation. The evaluation measures are listed in the ItemPredictionMeasures property. Additionally, 'num_users' and 'num_items' report the number of users that were used to compute the results and the number of items that were taken into account. Literature: C. Manning, P. Raghavan, H. Schütze: Introduction to Information Retrieval, Cambridge University Press, 2008

public static Evaluate ( IItemRecommender recommender, IPosOnlyFeedback test, IPosOnlyFeedback train, ICollection relevant_users, ICollection relevant_items ) : double>.Dictionary
recommender	IItemRecommender	item recommender
test	IPosOnlyFeedback	test cases
train	IPosOnlyFeedback	training data
relevant_users	ICollection	a collection of integers with all relevant users
relevant_items	ICollection	a collection of integers with all relevant items
return	double>.Dictionary

EvaluateOnline() public static method

Online evaluation for rankings of items

public static EvaluateOnline ( IItemRecommender recommender, IPosOnlyFeedback test, IPosOnlyFeedback train, ICollection relevant_users, ICollection relevant_items ) : double>.Dictionary
recommender	IItemRecommender	item recommender
test	IPosOnlyFeedback	test cases
train	IPosOnlyFeedback	training data (must be connected to the recommender's training data)
relevant_users	ICollection	a collection of integers with all relevant users
relevant_items	ICollection	a collection of integers with all relevant items
return	double>.Dictionary

MAP() public static method

Compute the mean average precision (MAP) of a list of ranked items

public static MAP ( int ranked_items, ICollection correct_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
return	double

MAP() public static method

Compute the mean average precision (MAP) of a list of ranked items

public static MAP ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
ignore_items	ICollection	a collection of item IDs which should be ignored for the evaluation
return	double

NDCG() public static method

Compute the normalized discounted cumulative gain (NDCG) of a list of ranked items

public static NDCG ( int ranked_items, ICollection correct_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
return	double

NDCG() public static method

Compute the normalized discounted cumulative gain (NDCG) of a list of ranked items

public static NDCG ( int ranked_items, ICollection correct_items, ICollection ignore_items ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
ignore_items	ICollection	a collection of item IDs which should be ignored for the evaluation
return	double

PrecisionAt() public static method

Compute the precision@N of a list of ranked items

public static PrecisionAt ( int ranked_items, ICollection correct_items, ICollection ignore_items, int n ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
ignore_items	ICollection	a collection of item IDs which should be ignored for the evaluation
n	int	the cutoff position in the list
return	double

PrecisionAt() public static method

Compute the precision@N of a list of ranked items

public static PrecisionAt ( int ranked_items, ICollection correct_items, int n ) : double
ranked_items	int	a list of ranked item IDs, the highest-ranking item first
correct_items	ICollection	a collection of positive/correct item IDs
n	int	the cutoff position in the list
return	double