DRecPy.Recommender.Baseline package

DRecPy.Recommender.Baseline.base_knn module

class DRecPy.Recommender.Baseline.base_knn.BaseKNN(k=20, m=5, sim_metric='adjusted_cosine', aggregation='weighted_mean', shrinkage=100, use_averages=False, **kwds)

Bases: DRecPy.Recommender.recommender_abc.RecommenderABC, abc.ABC

Base Collaborative Filtering recommender abstract class.

This class implements the skeleton methods for building a basic neighbour-based CF. The following methods are still required to be implemented: _fit(): should fit the model. _predict_default(): should return the default prediction value that is used when a minimum number of neighbours is not found. Only used when use_averages=True.

k

An integer representing the number of neighbours used to make a prediction. Default: 20.

m

An integer representing the minimum number of co-rated users/items required to validate the similarity value (if not valid, sim. value is set to 0). Default: 5.

sim_metric

Optional string representing the name of the similarity metric to use. Supported: ‘adjusted_cosine’, ‘cosine’, ‘cosine_cf’, ‘jaccard’, ‘msd’, ‘pearson’. Default: ‘adjusted_cosine’.

aggregation

Optional string representing the name of the aggregation approach to use. Supported: ‘mean’, ‘weighted_mean’. Default: ‘weighted_mean’.

shrinkage

Optional integer representing the discounting factor for computing the similarity between items / users (discounts less when #co-ratings increases). Default: 100.

use_averages

Optional boolean indicating whether to use item (for UserKNN) or user (for ItemKNN) averages when no neighbours are found. Default: True.

DRecPy.Recommender.Baseline.item_knn module

class DRecPy.Recommender.Baseline.item_knn.ItemKNN(**kwds)

Bases: DRecPy.Recommender.Baseline.base_knn.BaseKNN

Item-based KNN Collaborative Filtering recommender model.

Implementation of a basic item neighbour-based CF.

Public Methods:

fit(), predict(), recommend(), rank().

Attributes: See parent object BaseKNN obj:DRecPy.Recommender.Baseline.BaseKNN

DRecPy.Recommender.Baseline.user_knn module

class DRecPy.Recommender.Baseline.user_knn.UserKNN(**kwds)

Bases: DRecPy.Recommender.Baseline.base_knn.BaseKNN

User-based KNN Collaborative Filtering recommender model.

Implementation of a basic user neighbour-based CF.

Public Methods:

fit(), predict(), recommend(), rank().

Attributes: See parent object BaseKNN