Coverage for rulekit/params.py: 92%
88 statements
« prev ^ index » next coverage.py v7.6.10, created at 2025-01-07 11:26 +0000
1"""Contains constants and classes for specyfing models parameters
2"""
3from enum import Enum
4from typing import Callable
5from typing import Optional
6from typing import Tuple
7from typing import Union
9from jpype import JImplements
10from jpype import JOverride
11from jpype.types import JDouble
12from pydantic import BaseModel # pylint: disable=no-name-in-module
14MAX_INT: int = 2147483647 # max integer value in Java
16_UserDefinedMeasure = Callable[[float, float, float, float], float]
19def _user_defined_measure_factory(measure_function: _UserDefinedMeasure):
20 from adaa.analytics.rules.logic.quality import \
21 IUserMeasure # pylint: disable=import-outside-toplevel,import-error
23 @JImplements(IUserMeasure)
24 class _UserMeasure: # pylint: disable=invalid-name,missing-function-docstring
26 @JOverride
27 def getResult(self, p: JDouble, n: JDouble, P: JDouble, N: JDouble) -> float:
28 return measure_function(float(p), float(n), float(P), float(N))
30 return _UserMeasure()
33class Measures(Enum):
34 # pylint: disable=invalid-name
35 """Enum for different measures used during induction, pruning and voting.
37 You can ream more about each measure and its implementation
38 #41-rule-quality>`_ .
39 `here <https://github.com/adaa-polsl/RuleKit/wiki/4-Quality-and-evaluation
40 """
41 Accuracy = "Accuracy"
42 BinaryEntropy = "BinaryEntropy"
43 C1 = "C1"
44 C2 = "C2"
45 CFoil = "CFoil"
46 CN2Significnce = "CN2Significnce"
47 Correlation = "Correlation"
48 Coverage = "Coverage"
49 FBayesianConfirmation = "FBayesianConfirmation"
50 FMeasure = "FMeasure"
51 FullCoverage = "FullCoverage"
52 GeoRSS = "GeoRSS"
53 GMeasure = "GMeasure"
54 InformationGain = "InformationGain"
55 JMeasure = "JMeasure"
56 Kappa = "Kappa"
57 Klosgen = "Klosgen"
58 Laplace = "Laplace"
59 Lift = "Lift"
60 LogicalSufficiency = "LogicalSufficiency"
61 LogRank = "LogRank"
62 MEstimate = "MEstimate"
63 MutualSupport = "MutualSupport"
64 Novelty = "Novelty"
65 OddsRatio = "OddsRatio"
66 OneWaySupport = "OneWaySupport"
67 PawlakDependencyFactor = "PawlakDependencyFactor"
68 Precision = "Precision"
69 Q2 = "Q2"
70 RelativeRisk = "RelativeRisk"
71 Ripper = "Ripper"
72 RSS = "RSS"
73 RuleInterest = "RuleInterest"
74 SBayesian = "SBayesian"
75 Sensitivity = "Sensitivity"
76 Specificity = "Specificity"
77 TwoWaySupport = "TwoWaySupport"
78 WeightedLaplace = "WeightedLaplace"
79 WeightedRelativeAccuracy = "WeightedRelativeAccuracy"
80 YAILS = "YAILS"
83DEFAULT_PARAMS_VALUE = {
84 "minsupp_new": 0.05,
85 "induction_measure": Measures.Correlation,
86 "pruning_measure": Measures.Correlation,
87 "voting_measure": Measures.Correlation,
88 "max_growing": 0.0,
89 "enable_pruning": True,
90 "ignore_missing": False,
91 "max_uncovered_fraction": 0.0,
92 "select_best_candidate": False,
93 "complementary_conditions": False,
94 "control_apriori_precision": True,
95 "max_rule_count": 0,
96 "approximate_induction": False,
97 "approximate_bins_count": 100,
98 "mean_based_regression": True,
99 "extend_using_preferred": False,
100 "extend_using_automatic": False,
101 "induce_using_preferred": False,
102 "induce_using_automatic": False,
103 "consider_other_classes": False,
104 "preferred_conditions_per_rule": MAX_INT,
105 "preferred_attributes_per_rule": MAX_INT,
106 # Contrast sets
107 "minsupp_all": (0.8, 0.5, 0.2, 0.1),
108 "max_neg2pos": 0.5,
109 "max_passes_count": 5,
110 "penalty_strength": 0.5,
111 "penalty_saturation": 0.2,
112}
114_QualityMeasure = Union[Measures, _UserDefinedMeasure]
117class ModelsParams(BaseModel):
118 """Model for validating models hyperparameters"""
120 minsupp_new: Optional[float] = DEFAULT_PARAMS_VALUE["minsupp_new"]
121 induction_measure: Optional[_QualityMeasure] = DEFAULT_PARAMS_VALUE[
122 "induction_measure"
123 ]
124 pruning_measure: Optional[_QualityMeasure] = DEFAULT_PARAMS_VALUE["pruning_measure"]
125 voting_measure: Optional[_QualityMeasure] = DEFAULT_PARAMS_VALUE["voting_measure"]
126 max_growing: Optional[float] = DEFAULT_PARAMS_VALUE["max_growing"]
127 enable_pruning: Optional[bool] = DEFAULT_PARAMS_VALUE["enable_pruning"]
128 ignore_missing: Optional[bool] = DEFAULT_PARAMS_VALUE["ignore_missing"]
129 max_uncovered_fraction: Optional[float] = DEFAULT_PARAMS_VALUE[
130 "max_uncovered_fraction"
131 ]
132 select_best_candidate: Optional[bool] = DEFAULT_PARAMS_VALUE[
133 "select_best_candidate"
134 ]
135 complementary_conditions: Optional[bool] = DEFAULT_PARAMS_VALUE[
136 "complementary_conditions"
137 ]
138 max_rule_count: int = DEFAULT_PARAMS_VALUE["max_rule_count"]
141class ExpertModelParams(ModelsParams):
142 """Model for validating expert models hyperparameters"""
144 extend_using_preferred: Optional[bool] = DEFAULT_PARAMS_VALUE[
145 "extend_using_preferred"
146 ]
147 extend_using_automatic: Optional[bool] = DEFAULT_PARAMS_VALUE[
148 "extend_using_automatic"
149 ]
150 induce_using_preferred: Optional[bool] = DEFAULT_PARAMS_VALUE[
151 "induce_using_preferred"
152 ]
153 induce_using_automatic: Optional[bool] = DEFAULT_PARAMS_VALUE[
154 "induce_using_automatic"
155 ]
156 consider_other_classes: Optional[bool] = DEFAULT_PARAMS_VALUE[
157 "consider_other_classes"
158 ]
159 preferred_conditions_per_rule: Optional[int] = DEFAULT_PARAMS_VALUE[
160 "preferred_conditions_per_rule"
161 ]
162 preferred_attributes_per_rule: Optional[int] = DEFAULT_PARAMS_VALUE[
163 "preferred_attributes_per_rule"
164 ]
167class ContrastSetModelParams(ModelsParams):
168 """Model for validating contrast set models hyperparameters"""
170 minsupp_all: Tuple[float, float, float, float] = DEFAULT_PARAMS_VALUE["minsupp_all"]
171 max_neg2pos: Optional[float] = DEFAULT_PARAMS_VALUE["max_neg2pos"]
172 max_passes_count: Optional[int] = DEFAULT_PARAMS_VALUE["max_passes_count"]
173 penalty_strength: Optional[float] = DEFAULT_PARAMS_VALUE["penalty_strength"]
174 penalty_saturation: Optional[float] = DEFAULT_PARAMS_VALUE["penalty_saturation"]