teacher.tree

The teacher.tree implements the Decision Trees used as a white-box for the different explainers. Currently, there are two trees implemented as well as a representation for the rules.

The classes are the following ones:

BaseDecisionTree

Base class with the common methods that will be used for the rest of the decision trees.

ID3

Fuzzy adaptation of the ID3 algorithm that implements an inference that allows for an instance to traverse multiple branches of the tree. Used mainly as a baseline for comparison against other algorithms.

FDT

Multiway Fuzzy Decision Tree implemented as described here. This tree takes a set of fuzzy variables and a dataset and generates a decision tree, to get later a classification based on different types of inference.

---

class teacher.tree.BaseDecisionTree(features, th=0.0001, max_depth=2, min_num_examples=1, prunning=True)

Base abstract decision tree that provides the basic methods to implement the rest of the decision trees.

__init__(features, th=0.0001, max_depth=2, min_num_examples=1, prunning=True)

Parameters:

• features (list) – Sorted list of features as they will appear in the dataset

• th (float, optional) – Minimum gain threshold to keep branching the tree, by default 0.0001

• max_depth (int, optional) – Maximum depth of the tree, by default 2

• min_num_examples (int, optional) – Minimum number of examples per leaf, by default 1

• prunning (bool, optional) – Whether or not to prune the tree, by default True

abstract fit(X, y)

Build a decision tree classifier from the training set (X, y)

Parameters:

• X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

• y (array-like of shape (n_samples,)) – The target values (class labels) as integers or strings.

predict(X)

Predict class value for X.

The predicted class for each sample in X is returned.

Parameters:

X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

Returns:

y – The predicted classes

Return type:

array-like of shape (n_samples,)

score(X, y)

Return the mean accuracy on the given test data and labels

Parameters:

• X (array-like of shape (n_samples, n_features)) – Test samples.

• y (array-like of shape (n_samples,)) – True labels for X

Returns:

score – Mean accuracy of self.predict(X) wrt. y

Return type:

float

to_rule_based_system(verbose=False, simplify=False)

Return the tree as a rule-based system

Parameters:

• verbose (bool, optional) – debug flag, by default False

• simplify (bool, optional) – Whether or not to simplify the rules, by default False

Returns:

rule_system – List of the rules extracted from the tree

Return type:

list[Rule]

class teacher.tree.ID3(features, th=0.0001, max_depth=2, min_num_examples=1, prunning=True)

__init__(features, th=0.0001, max_depth=2, min_num_examples=1, prunning=True)

Parameters:

• features (list) – Sorted list of features as they will appear in the dataset

• th (float, optional) – Minimum gain threshold to keep branching the tree, by default 0.0001

• max_depth (int, optional) – Maximum depth of the tree, by default 2

• min_num_examples (int, optional) – Minimum number of examples per leaf, by default 1

• prunning (bool, optional) – Whether or not to prune the tree, by default True

fit(X, y, debug=False)

Build a decision tree classifier from the training set (X, y)

Parameters:

• X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

• y (array-like of shape (n_samples,)) – The target values (class labels) as integers or strings.

predict(X)

Predict class value for X.

The predicted class for each sample in X is returned.

Parameters:

X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

Returns:

y – The predicted classes

Return type:

array-like of shape (n_samples,)

score(X, y)

Return the mean accuracy on the given test data and labels

Parameters:

• X (array-like of shape (n_samples, n_features)) – Test samples.

• y (array-like of shape (n_samples,)) – True labels for X

Returns:

score – Mean accuracy of self.predict(X) wrt. y

Return type:

float

to_rule_based_system(verbose=False, simplify=False)

Return the tree as a rule-based system

Parameters:

• verbose (bool, optional) – debug flag, by default False

• simplify (bool, optional) – Whether or not to simplify the rules, by default False

Returns:

rule_system – List of the rules extracted from the tree

Return type:

list[Rule]

class teacher.tree.FDT(fuzzy_variables, fuzzy_threshold=0.0001, th=0.0001, max_depth=10, min_num_examples=1, prunning=True, t_norm=<ufunc 'minimum'>, voting='agg_vote')

__init__(fuzzy_variables, fuzzy_threshold=0.0001, th=0.0001, max_depth=10, min_num_examples=1, prunning=True, t_norm=<ufunc 'minimum'>, voting='agg_vote')

Parameters:

• fuzzy_variables (list[FuzzyVariable]) – List of the fuzzy variables used in the tree

• fuzzy_threshold (float, optional) – Minimum threshold for a pertenence degree to activate a fuzzy set, by default 0.0001

• th (float, optional) – Minimum gain threshold to keep branching the tree, by default 0.0001

• max_depth (int, optional) – Maximum depth of the tree, by default 2

• min_num_examples (int, optional) – Minimum number of examples per leaf, by default 1

• prunning (bool, optional) – Whether or not to prune the tree, by default True

• t_norm (function, optional) – function to be used as T-norm, by default np.minimum

• voting (str, optional) – {‘agg_vote’, ‘max_match’}, method of voting for the inference, by default ‘agg_vote’

fit(X, y)

Build a decision tree classifier from the training set (X, y)

Parameters:

• X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

• y (array-like of shape (n_samples,)) – The target values (class labels) as integers or strings.

predict(X)

Predict class value for X.

The predicted class for each sample in X is returned.

Parameters:

X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

Returns:

y – The predicted classes

Return type:

array-like of shape (n_samples,)

score(X, y)

Return the mean accuracy on the given test data and labels

Parameters:

• X (array-like of shape (n_samples, n_features)) – Test samples.

• y (array-like of shape (n_samples,)) – True labels for X

Returns:

score – Mean accuracy of self.predict(X) wrt. y

Return type:

float

to_rule_based_system(verbose=False, simplify=False)

Return the tree as a rule-based system

Parameters:

• verbose (bool, optional) – debug flag, by default False

• simplify (bool, optional) – Whether or not to simplify the rules, by default False

Returns:

rule_system – List of the rules extracted from the tree

Return type:

list[Rule]

class teacher.tree.FBDT(fuzzy_variables, fuzzy_threshold=0.0001, th=0.0001, max_depth=10, min_num_examples=1, prunning=True, t_norm=<ufunc 'minimum'>, voting='agg_vote')

__init__(fuzzy_variables, fuzzy_threshold=0.0001, th=0.0001, max_depth=10, min_num_examples=1, prunning=True, t_norm=<ufunc 'minimum'>, voting='agg_vote')

Parameters:

• fuzzy_variables (list[FuzzyVariable]) – List of the fuzzy variables used in the tree

• fuzzy_threshold (float, optional) – Minimum threshold for a pertenence degree to activate a fuzzy set, by default 0.0001

• th (float, optional) – Minimum gain threshold to keep branching the tree, by default 0.0001

• max_depth (int, optional) – Maximum depth of the tree, by default 2

• min_num_examples (int, optional) – Minimum number of examples per leaf, by default 1

• prunning (bool, optional) – Whether or not to prune the tree, by default True

• t_norm (function, optional) – function to be used as T-norm, by default np.minimum

• voting (str, optional) – {‘agg_vote’, ‘max_match’}, method of voting for the inference, by default ‘agg_vote’

fit(X, y)

Build a decision tree classifier from the training set (X, y)

Parameters:

• X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

• y (array-like of shape (n_samples,)) – The target values (class labels) as integers or strings.

predict(X)

Predict class value for X.

The predicted class for each sample in X is returned.

Parameters:

X (array-like of shape (n_samples, n_features)) – The training input samples. Internally, it will be converted to dtype=np.float64 if possible or dtype=object otherwise.

Returns:

y – The predicted classes

Return type:

array-like of shape (n_samples,)

score(X, y)

Return the mean accuracy on the given test data and labels

Parameters:

• X (array-like of shape (n_samples, n_features)) – Test samples.

• y (array-like of shape (n_samples,)) – True labels for X

Returns:

score – Mean accuracy of self.predict(X) wrt. y

Return type:

float

to_rule_based_system(verbose=False, simplify=False)

Return the tree as a rule-based system

Parameters:

• verbose (bool, optional) – debug flag, by default False

• simplify (bool, optional) – Whether or not to simplify the rules, by default False

Returns:

rule_system – List of the rules extracted from the tree

Return type:

list[Rule]

class teacher.tree.Rule(antecedent, consequent, weight, simplify=False, multiple_antecedents=False)

static map_rule_variables(rule, origin_fuzzy_variables, dest_fuzzy_variables, map_function='intersection')

Changes the fuzzy variables of the rule for ones that are defined in the same universe

Parameters:

• rule (Rule) – Original rule to map to the new variables

• origin_fuzzy_variables (list[FuzzyVariable]) – List with the original fuzzy variables

• dest_fuzzy_variables (list[FuzzyVariable]) – List with the destination fuzzy variables

• map_function (str, {'intersection', 'simmilarity'}) – Method to check the best fuzzy set to change

Returns:

Rule with the new variables

Return type:

Rule

Raises:

ValueError – If the universes of the variables are not the same it raises an error

matching(instance_membership, t_norm=<built-in function min>)

Matching that an instance has with the rule If there is some feature or value not existing in the instance, its matching degree is zero

Parameters:

• instance_membership (dict) – Membership of the instance with the format {feature: {value: pertenence degree}}

• t_norm (function, optional) – Operation to use as tnorm to get the matching, by default min

simplify()

If there are repeated features in the antecedent, it simplifies the rule

to_crisp(alpha_cut, fuzzy_variables)

Transform the rule to a crisp rule

Parameters:

alpha_cut (float) – Alpha cut to use to transform the rule

Returns:

Crisp rule

Return type:

Rule

to_json(fuzzy_variables)

Transform the rule to a json format

Parameters:

fuzzy_variables (list[FuzzyVariable]) – List with the fuzzy variables of the problem

Returns:

Json with the rule

Return type:

dict

static weighted_vote(rule_list, instance_membership)

Use the weighted vote inference method to return the consequent associated to an instance and a rule list given the instance membership

Parameters:

• rule_list (list[Rule]) – List with the rules that will be taken into account for the weighted vote method

• instance_membership (dict) – Membership of the instance with the format {feature: {value: pertenence degree}}

Returns:

consequent associated with the instance and the rule list

Return type:

string or number