EvoMSA.utils

class LabelEncoderWrapper

Wrapper of LabelEncoder. The idea is to keep the order when the classes are numbers at some point this will help improve the performance in ordinary classification problems

Parameters:

classifier (bool) – Specifies whether it is a classification problem

__init__(classifier=True)

property classifier

Whether EvoMSA is acting as classifier

fit(y)

Fit the label encoder

Parameters:

y (list or np.array) – Independent variables

Return type:

self

class Cache

Store the output of the text models

__init__(basename)

linearSVC_array(classifiers)

Transform LinearSVC into weight stored in array.array

Parameters:

classifers (list) – List of LinearSVC where each element is binary

bootstrap_confidence_interval(y: ~numpy.ndarray, hy: ~numpy.ndarray, metric: ~typing.Callable[[float, float], float] = <function <lambda>>, alpha: float = 0.05, nbootstrap: int = 500) → Tuple[float, float]

Confidence interval from predictions

class ConfidenceInterval

Estimate the confidence interval

>>> from EvoMSA import base
>>> from EvoMSA.utils import ConfidenceInterval
>>> from microtc.utils import tweet_iterator
>>> import os
>>> tweets = os.path.join(os.path.dirname(base.__file__), 'tests', 'tweets.json')
>>> D = list(tweet_iterator(tweets))
>>> X = [x['text'] for x in D]
>>> y = [x['klass'] for x in D]
>>> kw = dict(stacked_method="sklearn.naive_bayes.GaussianNB") 
>>> ci = ConfidenceInterval(X, y, evomsa_kwargs=kw)
>>> result = ci.estimate()

__init__(X: List[str], y: ndarray | list, Xtest: List[str] = None, y_test: ndarray | list = None, evomsa_kwargs: Dict = {}, folds: None | BaseCrossValidator = None) → None

class Linear

>>> from EvoMSA.model import Linear
>>> linear = Linear(coef=[12, 3], intercept=0.5, labels=[0, 'P'])
>>> X = np.array([[2, -1]])
>>> linear.decision_function(X)
21.5
>>> linear.predict(X)[0]
'P'

__init__(coef: list | ndarray, intercept: float = 0, labels: list | ndarray | None = None, N: int = 0) → None

property N

Size

property coef

Coefficients

property intercept

Bias or intercept

property labels

Classes

emoji_information(lang='es')

Download and load the Emoji statistics

Parameters:

lang (str) – [‘ar’, ‘zh’, ‘en’, ‘fr’, ‘pt’, ‘ru’, ‘es’]

>>> from EvoMSA.utils import emoji_information
>>> info = emoji_information()
>>> info['💧']
{'recall': 0.10575916230366492, 'ratio': 0.0003977123419509893, 'number': 3905}

load_dataset(lang='es', name='HA', k=None, d=17, func='most_common_by_type', v1=False)

Download and load the Dataset representation

Parameters:

• lang (str) – [‘ar’, ‘zh’, ‘en’, ‘es’]

• emoji (int) – emoji identifier

>>> from EvoMSA.utils import load_dataset, load_bow
>>> bow = load_bow(lang='en')
>>> ds = load_dataset(lang='en', name='travel', k=0)
>>> X = bow.transform(['this is funny'])
>>> df = ds.decision_function(X)

dataset_information(lang='es')

Download and load datasets information

Parameters:

lang (str) – [‘ar’, ‘zh’, ‘en’, ‘es’]

>>> from EvoMSA.utils import emoji_information
>>> info = dataset_information()