mojolearn.LocalOutlierFactor#

class mojolearn.LocalOutlierFactor(n_neighbors=20, *, algorithm='auto', leaf_size=30, metric='minkowski', p=2, metric_params=None, contamination='auto', novelty=False, n_jobs=None)[source]#

Bases: _XNeighbors

Unsupervised outlier detection by the local outlier factor (LOF).

Scores each sample by how isolated it is relative to its k nearest neighbors, following scikit-learn’s LocalOutlierFactor. Neighbor searches are exact brute force with the euclidean metric; ties between equal distances go to the lower training index, and a duplicate of a training row counts as its neighbor. Computation is float32 (scikit-learn uses float64).

With novelty=False (the default) use fit_predict() to label the training data. With novelty=True fit on clean data and call predict(), decision_function() or score_samples() on new data.

Parameters:

  • n_neighbors (int, default 20) – Number of neighbors. fit uses min(n_neighbors, n_samples - 1).

  • algorithm (str, default ‘auto’) – Accepted for scikit-learn compatibility; every search is exact brute force.

  • leaf_size (int, default 30) – Accepted for scikit-learn compatibility; unused.

  • metric (str, default ‘minkowski’) – 'euclidean', or 'minkowski' with p=2. Any other metric raises NotImplementedError.

  • p (int, default 2) – Minkowski power; only 2 is supported.

  • metric_params (dict or None, default None) – Not supported; a non-empty value raises NotImplementedError.

  • contamination (‘auto’ or float, default ‘auto’) – Expected outlier fraction, in (0, 0.5]. 'auto' sets offset_ to -1.5; a float sets it to that percentile of the training scores.

  • novelty (bool, default False) – Whether the model scores new data (predict, decision_function, score_samples) instead of the training data (fit_predict).

  • n_jobs (int or None, default None) – Accepted for scikit-learn compatibility; unused.

  • numeric_mode ({‘identical’, ‘deterministic’, ‘fast’} or None, default None) – Numeric mode for this estimator ('fast' is Apple only). None uses the process default.

Variables:

  • negative_outlier_factor_ (Array of shape (n_samples,)) – Negative LOF of each training sample; lower is more abnormal.

  • offset_ (float) – Threshold on the score; samples below it are outliers.

  • n_neighbors_ (int) – Number of neighbors actually used.

  • n_features_in_ (int) – Number of features seen in fit.

  • n_samples_fit_ (int) – Number of training samples.

Examples

>>> import numpy as np, mojolearn
>>> X = np.random.default_rng(0).random((1000, 8), dtype=np.float32)
>>> labels = mojolearn.LocalOutlierFactor(n_neighbors=20).fit_predict(X)
__init__(n_neighbors=20, *, algorithm='auto', leaf_size=30, metric='minkowski', p=2, metric_params=None, contamination='auto', novelty=False, n_jobs=None)[source]#

Methods

decision_function(X)

Shifted outlier score of new samples: score_samples(X) - offset_.

fit(X[, y])

Fit the model on the training data.

fit_predict(X[, y])

Fit the model and label the training samples.

get_params([deep])

numeric_mode_used()

The numeric mode this estimator runs in.

predict(X)

Label new samples as inliers or outliers.

score_samples(X)

Negative local outlier factor of new samples.

set_params(**params)

Set constructor parameters.

vendor_used()

The GPU backend this estimator runs on.
fit(X, y=None)[source]#

Fit the model on the training data.

Parameters:

  • X (array-like or scipy.sparse matrix of shape (n_samples, n_features)) – Training data; at least 2 samples. Sparse input is densified.

  • y (None) – Ignored.

Returns:

self (LocalOutlierFactor) – The fitted estimator.

fit_predict(X, y=None)[source]#

Fit the model and label the training samples.

Available only with novelty=False; otherwise raises AttributeError.

Parameters:

  • X (array-like of shape (n_samples, n_features)) – Training data.

  • y (None) – Ignored.

Returns:

labels (Array of shape (n_samples,), int64) – -1 for outliers, 1 for inliers.

score_samples(X)[source]#

Negative local outlier factor of new samples.

Available only with novelty=True; otherwise raises AttributeError. Lower scores are more abnormal.

Parameters:

X (array-like of shape (n_queries, n_features)) – New samples.

Returns:

scores (Array of shape (n_queries,), float32)

decision_function(X)[source]#

Shifted outlier score of new samples: score_samples(X) - offset_.

Available only with novelty=True. Negative values are outliers.

Parameters:

X (array-like of shape (n_queries, n_features)) – New samples.

Returns:

scores (Array of shape (n_queries,), float32)

predict(X)[source]#

Label new samples as inliers or outliers.

Available only with novelty=True; otherwise raises AttributeError.

Parameters:

X (array-like of shape (n_queries, n_features)) – New samples.

Returns:

labels (Array of shape (n_queries,), int64) – -1 for outliers, 1 for inliers.