Coverage for src/sensai/sklearn/sklearn_classification.py: 71%

1import logging 

2from typing import Union, Optional 

3 

4import numpy as np 

5import sklearn.ensemble 

6import sklearn.naive_bayes 

7import sklearn.neural_network 

8import sklearn.tree 

9from sklearn.ensemble import RandomForestClassifier 

10from sklearn.tree import DecisionTreeClassifier 

11 

12from .sklearn_base import AbstractSkLearnVectorClassificationModel, FeatureImportanceProviderSkLearnClassification 

13 

14log = logging.getLogger(__name__) 

15 

16 

17class SkLearnDecisionTreeVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

18 def __init__(self, min_samples_leaf=1, random_state=42, **model_args): 

19 super().__init__(DecisionTreeClassifier, 

20 min_samples_leaf=min_samples_leaf, random_state=random_state, **model_args) 

21 

22 def is_sample_weight_supported(self) -> bool: 

23 return True 

24 

25 

26class SkLearnRandomForestVectorClassificationModel(AbstractSkLearnVectorClassificationModel, 

27 FeatureImportanceProviderSkLearnClassification): 

28 def __init__(self, n_estimators=100, min_samples_leaf=1, random_state=42, use_balanced_class_weights=False, **model_args): 

29 super().__init__(RandomForestClassifier, 

30 random_state=random_state, min_samples_leaf=min_samples_leaf, n_estimators=n_estimators, 

31 use_balanced_class_weights=use_balanced_class_weights, 

32 **model_args) 

33 

34 def is_sample_weight_supported(self) -> bool: 

35 return True 

36 

37 

38class SkLearnMLPVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

39 def __init__(self, hidden_layer_sizes=(100,), activation: str = "relu", 

40 solver: str = "adam", batch_size: Union[int, str] = "auto", random_state: Optional[int] = 42, 

41 max_iter: int = 200, early_stopping: bool = False, n_iter_no_change: int = 10, **model_args): 

42 """ 

43 :param hidden_layer_sizes: the sequence of hidden layer sizes 

44 :param activation: {"identity", "logistic", "tanh", "relu"} the activation function to use for hidden layers (the one used for the 

45 output layer is always 'identity') 

46 :param solver: {"adam", "lbfgs", "sgd"} the name of the solver to apply 

47 :param batch_size: the batch size or "auto" for min(200, data set size) 

48 :param random_state: the random seed for reproducability; use None if it shall not be specifically defined 

49 :param max_iter: the number of iterations (gradient steps for L-BFGS, epochs for other solvers) 

50 :param early_stopping: whether to use early stopping (stop training after n_iter_no_change epochs without improvement) 

51 :param n_iter_no_change: the number of iterations after which to stop early (if early_stopping is enabled) 

52 :param model_args: additional arguments to pass on to MLPClassifier, see 

53 https://scikit-learn.org/stable/modules/generated/sklearn.neural_network.MLPClassifier.html 

54 """ 

55 super().__init__(sklearn.neural_network.MLPClassifier, hidden_layer_sizes=hidden_layer_sizes, activation=activation, 

56 random_state=random_state, solver=solver, batch_size=batch_size, max_iter=max_iter, early_stopping=early_stopping, 

57 n_iter_no_change=n_iter_no_change, **model_args) 

58 

59 def is_sample_weight_supported(self) -> bool: 

60 return False 

61 

62 

63class SkLearnMultinomialNBVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

64 def __init__(self, **model_args): 

65 super().__init__(sklearn.naive_bayes.MultinomialNB, **model_args) 

66 

67 def is_sample_weight_supported(self) -> bool: 

68 return True 

69 

70 

71class SkLearnSVCVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

72 def __init__(self, random_state=42, **model_args): 

73 super().__init__(sklearn.svm.SVC, random_state=random_state, **model_args) 

74 

75 def is_sample_weight_supported(self) -> bool: 

76 return True 

77 

78 

79class SkLearnLogisticRegressionVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

80 def __init__(self, random_state=42, **model_args): 

81 super().__init__(sklearn.linear_model.LogisticRegression, random_state=random_state, **model_args) 

82 

83 def is_sample_weight_supported(self) -> bool: 

84 return True 

85 

86 

87class SkLearnKNeighborsVectorClassificationModel(AbstractSkLearnVectorClassificationModel): 

88 def __init__(self, **model_args): 

89 super().__init__(sklearn.neighbors.KNeighborsClassifier, **model_args) 

90 

91 def _predict_sklearn(self, input_values): 

92 # Apply a transformation to fix a bug in sklearn 1.3.0 (and perhaps earlier versions): 

93 # https://github.com/scikit-learn/scikit-learn/issues/26768 

94 inputs = np.ascontiguousarray(input_values) 

95 

96 return super()._predict_sklearn(inputs) 

97 

98 def is_sample_weight_supported(self) -> bool: 

99 return False