Learning classifiers
 |  |  |
import pyagrum.skbn as skbnimport pyagrum.lib.notebook as gnbskbn is a pyAgrum’s module that allows to use bayesian networks as classifier in the scikit-learn environment.
Initialization of parameters
Section titled “Initialization of parameters”First, we initialize the parameters to indicate properties we want our classifier to have.
BNTest = skbn.createBNClassifier( learningMethod="Chow-Liu", prior="Smoothing", priorWeight=0.5, discretizationStrategy="quantile", usePR=True, significant_digit=13,)Then, we train the classifier thanks to two types of objects.
Learn from csv file
Section titled “Learn from csv file”BNTest.fitFromTabular(data="res/creditCardTest.csv", targetName="Class")BNClassifier(learningMethod='Chow-Liu', prior='Smoothing', priorWeight=0.5,Parameters
Fitted attributes
type_processor <pyagrum.lib....t 0x1129e8980> learningMethod 'Chow-Liu' prior 'Smoothing' priorWeight 0.5 usePR True significant_digit 13 scoringType 'BIC' constraints None possibleSkeleton None DirichletCsv None beta 1 SPDX-License-Identifier: BSD-3-Clause */
function copyToClipboard(text, element) { // Get the parameter prefix from the closest toggleable content const toggleableContent = element.closest(‘.sk-toggleable__content’); const paramPrefix = toggleableContent ? toggleableContent.dataset.paramPrefix : ”; const fullParamName = paramPrefix ?
${paramPrefix}${text}: text;}
document.querySelectorAll(‘.copy-paste-icon’).forEach(function(element) { const toggleableContent = element.closest(‘.sk-toggleable__content’); const paramPrefix = toggleableContent ? toggleableContent.dataset.paramPrefix : ”;
});
/**
Copy the list of feature names formatted as a Python list.
@param {HTMLElement} element - The copy button inside a
.featuresblock; its siblingscontain a
detailselement and a table containing feature named.@returns {boolean} Always returns
falseso callers can prevent the default click behavior. */ function copyFeatureNamesToClipboard(element) { var detailsElem = element.closest(‘.features’).querySelector(‘details’); var wasOpen = detailsElem.open; detailsElem.open = true; var content = element.closest(‘.features’).querySelector(‘tbody’) .innerText.trim(); if (!wasOpen) detailsElem.open = false; const rows = content.split(‘\n’).map(row =>"${row}"); const formattedText =[\n${rows.join(',\n')},\n]; const originalHTML = element.innerHTML.replace(’✔’, ”); const originalStyle = element.style; const copyMark = document.createElement(‘span’); copyMark.innerHTML = ’✔’; copyMark.style.color = ‘blue’; copyMark.style.fontSize = ‘1em’;navigator.clipboard.writeText(formattedText) .then(() => { element.style.display = ‘none’; element.parentElement.appendChild(copyMark);
return false; } /**
Adapted from Skrub
@returns “light” or “dark” */ function detectTheme(element) { const body = document.querySelector(‘body’);
// Check VSCode theme const themeKindAttr = body.getAttribute(‘data-vscode-theme-kind’); const themeNameAttr = body.getAttribute(‘data-vscode-theme-name’);
if (themeKindAttr && themeNameAttr) { const themeKind = themeKindAttr.toLowerCase(); const themeName = themeNameAttr.toLowerCase();
}
// Check Jupyter theme if (body.getAttribute(‘data-jp-theme-light’) === ‘false’) { return ‘dark’; } else if (body.getAttribute(‘data-jp-theme-light’) === ‘true’) { return ‘light’; }
// Guess based on a parent element’s color const color = window.getComputedStyle(element.parentNode, null).getPropertyValue(‘color’); const match = color.match(/^rgb\s*(\s*(\d+)\s*,\s*(\d+)\s*,\s*(\d+)\s*)\s*$/i); if (match) { const [r, g, b] = [ parseFloat(match[1]), parseFloat(match[2]), parseFloat(match[3]) ];
}
// Fallback to system preference return window.matchMedia(‘(prefers-color-scheme: dark)‘).matches ? ‘dark’ : ‘light’; }
function forceTheme(elementId) { const estimatorElement = document.querySelector(
#${elementId}); if (estimatorElement === null) { console.error(Element with id ${elementId} not found.); } else { const theme = detectTheme(estimatorElement); estimatorElement.classList.add(theme); } }forceTheme(‘sk-container-id-1’);
Name Type Value MarkovBlanket_ BayesNet (pyagrum.Baye...: 9, mem: 96o} bn_ BayesNet (pyagrum.Baye...mem: 5Ko 776o} classes_ ndarray[float64](2,) [0.,1.] feature_names_in_ ndarray[object](30,) ['Time','V1','V2',...,'V27','V28','Amount'] fromModel_ bool False label_ str '1.0' learner_ BNLearner (pyagrum.BNLe... : 0.500000 n_features_in_ int 30 targetType_ Float64DType dtype('float64') target_ str 'Class' threshold_ float64 0.121 variableNameIndexDictionary_ dict {'Amount': 29, 'Time': 0, 'V1': 1, 'V10': 10, ...}
Learn from array-likes
Section titled “Learn from array-likes”We use a method to transform the csv file in two array-likes in order to train from the same database.
Parameters
Fitted attributes
type_processor <pyagrum.lib....t 0x112c022c0> prior 'Smoothing' priorWeight 0.5 usePR True significant_digit 13 learningMethod 'MIIC' scoringType 'BIC' constraints None possibleSkeleton None DirichletCsv None beta 1 SPDX-License-Identifier: BSD-3-Clause */
function copyToClipboard(text, element) { // Get the parameter prefix from the closest toggleable content const toggleableContent = element.closest(‘.sk-toggleable__content’); const paramPrefix = toggleableContent ? toggleableContent.dataset.paramPrefix : ”; const fullParamName = paramPrefix ?
${paramPrefix}${text}: text;}
document.querySelectorAll(‘.copy-paste-icon’).forEach(function(element) { const toggleableContent = element.closest(‘.sk-toggleable__content’); const paramPrefix = toggleableContent ? toggleableContent.dataset.paramPrefix : ”;
});
/**
Copy the list of feature names formatted as a Python list.
@param {HTMLElement} element - The copy button inside a
.featuresblock; its siblingscontain a
detailselement and a table containing feature named.@returns {boolean} Always returns
falseso callers can prevent the default click behavior. */ function copyFeatureNamesToClipboard(element) { var detailsElem = element.closest(‘.features’).querySelector(‘details’); var wasOpen = detailsElem.open; detailsElem.open = true; var content = element.closest(‘.features’).querySelector(‘tbody’) .innerText.trim(); if (!wasOpen) detailsElem.open = false; const rows = content.split(‘\n’).map(row =>"${row}"); const formattedText =[\n${rows.join(',\n')},\n]; const originalHTML = element.innerHTML.replace(’✔’, ”); const originalStyle = element.style; const copyMark = document.createElement(‘span’); copyMark.innerHTML = ’✔’; copyMark.style.color = ‘blue’; copyMark.style.fontSize = ‘1em’;navigator.clipboard.writeText(formattedText) .then(() => { element.style.display = ‘none’; element.parentElement.appendChild(copyMark);
return false; } /**
Adapted from Skrub
@returns “light” or “dark” */ function detectTheme(element) { const body = document.querySelector(‘body’);
// Check VSCode theme const themeKindAttr = body.getAttribute(‘data-vscode-theme-kind’); const themeNameAttr = body.getAttribute(‘data-vscode-theme-name’);
if (themeKindAttr && themeNameAttr) { const themeKind = themeKindAttr.toLowerCase(); const themeName = themeNameAttr.toLowerCase();
}
// Check Jupyter theme if (body.getAttribute(‘data-jp-theme-light’) === ‘false’) { return ‘dark’; } else if (body.getAttribute(‘data-jp-theme-light’) === ‘true’) { return ‘light’; }
// Guess based on a parent element’s color const color = window.getComputedStyle(element.parentNode, null).getPropertyValue(‘color’); const match = color.match(/^rgb\s*(\s*(\d+)\s*,\s*(\d+)\s*,\s*(\d+)\s*)\s*$/i); if (match) { const [r, g, b] = [ parseFloat(match[1]), parseFloat(match[2]), parseFloat(match[3]) ];
}
// Fallback to system preference return window.matchMedia(‘(prefers-color-scheme: dark)‘).matches ? ‘dark’ : ‘light’; }
function forceTheme(elementId) { const estimatorElement = document.querySelector(
#${elementId}); if (estimatorElement === null) { console.error(Element with id ${elementId} not found.); } else { const theme = detectTheme(estimatorElement); estimatorElement.classList.add(theme); } }forceTheme(‘sk-container-id-2’);
Create a classifier from a Bayesian network
Section titled “Create a classifier from a Bayesian network”If we already have a Bayesian network with learned parameters, we can create a classifier that uses it. In this case we do not have to train the classifier on data since it the Bayesian network is already trained.
Then, we work with functions from scikit-learn like score. We can also call it with a csv file or two array-likes.
Prediction for classifier
Section titled “Prediction for classifier”Prediction with csv file
Section titled “Prediction with csv file”Prediction with array-like
Section titled “Prediction with array-like”ROC and Precision-Recall curves with all methods
Section titled “ROC and Precision-Recall curves with all methods”In addition (and of course), we can work with functions from pyagrum (from
pyagrum.lib.bn2roc).
Name Type Value MarkovBlanket_ BayesNet (pyagrum.Baye...m: 293Ko 288o} bn_ BayesNet (pyagrum.Baye...Mo 117Ko 968o} classes_ ndarray[float64](2,) [0.,1.] feature_names_in_ ndarray[object](30,) ['Time','V1','V2',...,'V27','V28','Amount'] fromModel_ bool False label_ str '1.0' learner_ BNLearner (pyagrum.BNLe... : 0.500000 n_features_in_ int 30 targetType_ Float64DType dtype('float64') target_ str 'Class' threshold_ float64 0.5612 variableNameIndexDictionary_ dict {'Amount': 29, 'Time': 0, 'V1': 1, 'V10': 10, ...}
