{"id":87971,"date":"2021-09-10T10:31:25","date_gmt":"2021-09-10T09:31:25","guid":{"rendered":"https:\/\/datascientest.com\/?p=87971"},"modified":"2024-01-25T11:54:40","modified_gmt":"2024-01-25T11:54:40","slug":"isolation-forest","status":"publish","type":"post","link":"https:\/\/datascientest.com\/isolation-forest","title":{"rendered":"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"87971\" class=\"elementor elementor-87971\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-3d5f1be e-flex e-con-boxed e-con\" data-id=\"3d5f1be\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-056df9e elementor-widget elementor-widget-heading\" data-id=\"056df9e\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><h2 class=\"elementor-heading-title elementor-size-default\">Aujourd\u2019hui, nous nous int\u00e9ressons \u00e0 l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires tels que la d\u00e9tection de fraude ou le diagnostic d\u2019une maladie. Cette technique pr\u00e9sent\u00e9e \u00e0 l\u2019Eighth IEEE International Conference en 2008 est la premi\u00e8re technique de classification d\u00e9di\u00e9e \u00e0 la d\u00e9tection d\u2019anomalies bas\u00e9e sur l\u2019isolation.<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-facdcc8 e-flex e-con-boxed e-con\" data-id=\"facdcc8\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-b3277a4 elementor-widget elementor-widget-heading\" data-id=\"b3277a4\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Le probl\u00e8me de l\u2019Accuracy Paradox<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-01ba6af elementor-widget elementor-widget-text-editor\" data-id=\"01ba6af\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Lorsqu\u2019on fait face \u00e0 un probl\u00e8me de classification de classes en Machine Learning, il arrive parfois <a href=\"https:\/\/datascientest.com\/comment-gerer-les-problemes-de-classification-desequilibree-partie-i\">que le jeu de donn\u00e9es soit d\u00e9s\u00e9quilibr\u00e9<\/a>. Plus pr\u00e9cis\u00e9ment, on constate une proportion bien plus grande d\u2019une classe vis-\u00e0-vis de l\u2019autre au sein de l\u2019\u00e9chantillon d\u2019apprentissage. Ce probl\u00e8me est une caract\u00e9ristique fr\u00e9quente des probl\u00e8mes de classification binaires.\u00a0On pourrait en premi\u00e8re approche, penser \u00e0 appliquer un algorithme de classification classique, comme les <a href=\"https:\/\/datascientest.com\/random-forest-definition\">arbres de d\u00e9cisions<\/a>, les <a href=\"https:\/\/datascientest.com\/knn\">K-NN<\/a> ou les <a href=\"https:\/\/datascientest.com\/svm\">SVM<\/a> pr\u00e9sent\u00e9s sur notre blog. Toutefois, ces algorithmes ne sont en r\u00e9alit\u00e9 pas en mesure de g\u00e9rer ces jeux de donn\u00e9es atypiques o\u00f9 les disparit\u00e9s entre les classes sont assez fortes. Ils tendent \u00e0 maximiser par le biais de leur fonction de perte des quantit\u00e9s tel que <\/span><b>l\u2019accuracy<\/b><span style=\"font-weight: 400;\">, mais sans tenir compte de la distribution des donn\u00e9es.<\/span><\/p><p><span style=\"font-weight: 400;\">En pratique, cela se retrouve sur des mod\u00e8les qui, entra\u00een\u00e9s sur un <a href=\"https:\/\/datascientest.com\/dataset-definition\">dataset<\/a> domin\u00e9 par une classe, vont rendre compte lors de la phase d\u2019\u00e9valuation d\u2019un taux de pr\u00e9dictions correctes, appel\u00e9 \u00ab\u00a0<a href=\"https:\/\/datascientest.com\/danielcomment-lire-et-exploiter-une-matrice-de-confusion\">accuracy<\/a>\u00ab\u00a0, \u00e9lev\u00e9 mais ne seront pas pertinents \u00e0 l\u2019usage.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Prenons un exemple pour illustrer ce paradoxe. Consid\u00e9rons un probl\u00e8me de <\/span><b>d\u00e9tection de fraude bancaire<\/b><span style=\"font-weight: 400;\">, une banque cherche \u00e0 d\u00e9terminer parmi un grand nombre de transactions, quelles sont les transactions frauduleuses \u00e0 partir d\u2019un certain nombre de variables explicatives. Ces transactions frauduleuses repr\u00e9sentent environ 11% des transactions dans notre dataset. La classification \u00e0 l\u2019aide d\u2019un mod\u00e8le simple tel que le SVM <a href=\"https:\/\/datascientest.com\/tout-savoir-sur-scikit-learn\">\u00e0 l\u2019aide de scikit learn<\/a> donne par exemple donne le score suivant :<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1d65db1 elementor-widget elementor-widget-image\" data-id=\"1d65db1\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"261\" height=\"32\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/Ynrcj554OXGvMvJ9dVOKrIlut_xhW5AYb5_QaIXSE5EqzRP9Fnex4FSZjbBqnQxgzB2fV2A2A26jDdMY5vdFdFjcz40AU27eKozCzjHPmdheJJTkjU8k_9njuL1FbCvq1.jpg\" class=\"attachment-large size-large wp-image-87986\" alt=\"score\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d5f88fb elementor-widget elementor-widget-text-editor\" data-id=\"d5f88fb\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400; color: #000000;\">Toutefois en jetant un \u0153il \u00e0 <a href=\"https:\/\/datascientest.com\/matrice-de-confusion\">la matrice de confusion<\/a> on remarque vite que le comportement de l\u2019algorithme est assez na\u00eff :<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e622f6d elementor-widget elementor-widget-image\" data-id=\"e622f6d\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"171\" height=\"115\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/Classe-pre\u0301dite.jpg\" class=\"attachment-full size-full wp-image-87987\" alt=\"comportement de l\u2019algorithme\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-87c844c elementor-widget elementor-widget-text-editor\" data-id=\"87c844c\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">En effet, sur 219 transactions frauduleuses seules 25 ont bien \u00e9t\u00e9 identifi\u00e9es comme frauduleuses par notre mod\u00e8le.\u00a0 Nous constatons ainsi que le pouvoir de pr\u00e9diction du mod\u00e8le n\u2019est pas aussi puissant que ce que pourrait laisser croire l\u2019accuracy. C\u2019est une m\u00e9trique qui est trop biais\u00e9 par un d\u00e9s\u00e9quilibre des classes, ce ph\u00e9nom\u00e8ne est appel\u00e9<\/span><a href=\"https:\/\/towardsdatascience.com\/accuracy-paradox-897a69e2dd9b\"> l\u2019accuracy paradox<\/a><b>.<\/b><\/p><p><i><span style=\"font-weight: 400;\">Un mod\u00e8le avec un score d\u2019accuracy plus \u00e9lev\u00e9 peut avoir un pouvoir de pr\u00e9diction plus faible qu\u2019un mod\u00e8le avec une accuracy plus faible.<\/span><\/i><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-198b0a7 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"198b0a7\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formations-data\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Apprendre \u00e0 corriger l'accuracy paradox<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-dda8f1d e-flex e-con-boxed e-con\" data-id=\"dda8f1d\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-d87d120 elementor-widget elementor-widget-heading\" data-id=\"d87d120\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">L\u2019Isolation Forest : c\u2019est quoi ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1199705 elementor-widget elementor-widget-text-editor\" data-id=\"1199705\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Mais alors comment faire face \u00e0 ce probl\u00e8me qui vient ruiner les efforts de plusieurs algorithmes classiques rondement entra\u00een\u00e9s ? Il existe plusieurs solutions comme par exemple les m\u00e9thodes de r\u00e9-\u00e9chantillonnages comme <\/span><b>l&rsquo;undersampling<\/b><span style=\"font-weight: 400;\"> ou <\/span><b>l&rsquo;oversampling<\/b><span style=\"font-weight: 400;\"> qui s\u00e9lectionnent al\u00e9atoirement des donn\u00e9es afin d\u2019\u00e9quilibrer le ratio de classe. Toutefois, lorsque ces m\u00e9thodes ne sont pas assez efficaces, il est d\u2019usage de repenser le probl\u00e8me et de consid\u00e9rer la d\u00e9tection d\u2019anomalies comme un probl\u00e8me de <a href=\"https:\/\/datascientest.com\/apprentissage-non-supervise\">classification non supervis\u00e9e<\/a>. C\u2019est l\u00e0 qu\u2019intervient le tant attendu <\/span><b>Isolation Forest<\/b><span style=\"font-weight: 400;\"> !<\/span><\/p><p><span style=\"font-weight: 400;\">Dans Isolation Forest, on retrouve Isolation car c\u2019est une technique de d\u00e9tection d\u2019anomalies qui identifie <\/span><b>directement<\/b><span style=\"font-weight: 400;\"> les anomalies (commun\u00e9ment appel\u00e9es \u201c<\/span><b>outliers<\/b><span style=\"font-weight: 400;\">\u201d) contrairement aux techniques usuelles qui discriminent les points vis-\u00e0-vis d\u2019un profil global \u00ab normalis\u00e9 \u00bb.<\/span><\/p><p><span style=\"font-weight: 400;\">Le principe de cet algorithme est tr\u00e8s simple :<\/span><\/p><ul><li style=\"font-weight: 400;\" aria-level=\"1\"><span style=\"font-weight: 400;\">On s\u00e9lectionne une <\/span><b>variable<\/b><span style=\"font-weight: 400;\"> (feature) de fa\u00e7on <\/span><b>al\u00e9atoire<\/b><span style=\"font-weight: 400;\">.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><span style=\"font-weight: 400;\">Ensuite on r\u00e9alise un partitionnement al\u00e9atoire du jeu de donn\u00e9es selon cette variable de telle sorte \u00e0 obtenir <\/span><b>deux<\/b><span style=\"font-weight: 400;\"> sous-ensembles de donn\u00e9es.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><span style=\"font-weight: 400;\">On r\u00e9p\u00e8te les deux \u00e9tapes pr\u00e9c\u00e9dentes jusqu\u2019\u00e0 ce qu\u2019une donn\u00e9e soit <\/span><b>isol\u00e9e<\/b><span style=\"font-weight: 400;\">.<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><span style=\"font-weight: 400;\">De mani\u00e8re <\/span><b>r\u00e9cursive<\/b><span style=\"font-weight: 400;\">, on r\u00e9p\u00e8te les \u00e9tapes pr\u00e9c\u00e9dentes.<\/span><\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-4fd2444 elementor-widget elementor-widget-image\" data-id=\"4fd2444\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"500\" height=\"300\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/python_anomaly_detection_isolation_forest.gif\" class=\"attachment-large size-large wp-image-87990\" alt=\"\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">fonctionnement de l'arlgorithme<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-78b66b8 elementor-widget elementor-widget-text-editor\" data-id=\"78b66b8\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<span style=\"font-weight: 400;\">De la m\u00eame mani\u00e8re que les algorithmes ensemblistes tel que le <a href=\"https:\/\/datascientest.com\/random-forest-definition\">Random Forest<\/a>, on cr\u00e9e une for\u00eat (d\u2019o\u00f9 le nom d\u2019Isolation Forest) compos\u00e9e de dizaines ou centaines d\u2019arbres dont on va <\/span><b>combiner<\/b><span style=\"font-weight: 400;\"> les r\u00e9sultats afin d\u2019obtenir un meilleur r\u00e9sultat.<\/span>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-c9f73d4 elementor-widget elementor-widget-image\" data-id=\"c9f73d4\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"541\" height=\"388\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/z8olQRY0Nc9JS4lj0c0V7lWFxz9zhihuOLDX6YmrbxOktSHag3A2eXu-BNYOAD6_7yxtJUeL1vBtSOgX6TDy_LLcZCQF_4l9d8ZQQfgrABMjb2E1CAECCcvsF9PnSa6nY-1.jpg\" class=\"attachment-large size-large wp-image-87989\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/z8olQRY0Nc9JS4lj0c0V7lWFxz9zhihuOLDX6YmrbxOktSHag3A2eXu-BNYOAD6_7yxtJUeL1vBtSOgX6TDy_LLcZCQF_4l9d8ZQQfgrABMjb2E1CAECCcvsF9PnSa6nY-1.jpg 541w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/08\/z8olQRY0Nc9JS4lj0c0V7lWFxz9zhihuOLDX6YmrbxOktSHag3A2eXu-BNYOAD6_7yxtJUeL1vBtSOgX6TDy_LLcZCQF_4l9d8ZQQfgrABMjb2E1CAECCcvsF9PnSa6nY-1-300x215.jpg 300w\" sizes=\"(max-width: 541px) 100vw, 541px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">illustration de l'isolation forest<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-0ef4f2f elementor-widget elementor-widget-text-editor\" data-id=\"0ef4f2f\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Comment alors obtenir le r\u00e9sultat sur un arbre ? <\/span><b>C\u2019est-\u00e0-dire comment d\u00e9finir si un point est une anomalie ou non ? <\/b><span style=\"font-weight: 400;\">La construction de l\u2019algorithme donne lieu \u00e0 une structure d\u2019arbre, dont les n\u0153uds sont les ensembles partitionn\u00e9s durant les \u00e9tapes de l\u2019algorithme et les feuilles sont les points isol\u00e9s. Intuitivement les anomalies auront tendance \u00e0 \u00eatre les feuilles <\/span><b>les plus proches de la racine de l\u2019arbre<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">En fonction de la distance \u00e0 la racine,<\/span><b> on associe \u00e0 chaque feuille d\u2019un arbre un score d\u2019anomalie<\/b><span style=\"font-weight: 400;\">, on <\/span><b>moyenne <\/b><span style=\"font-weight: 400;\">ce score d\u2019anomalie sur l\u2019ensemble des arbres qui ont \u00e9t\u00e9 construits r\u00e9cursivement et on obtient un r\u00e9sultat final.\u00a0 On peut alors selon la situation<\/span><b> jouer sur les param\u00e8tres<\/b><span style=\"font-weight: 400;\"> tels que le nombre d\u2019arbres ou la proportion d\u2019anomalies \u00e0 consid\u00e9rer afin d\u2019obtenir un <\/span><b>r\u00e9sultat plus robuste<\/b><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">L\u2019Isolation Forest est tr\u00e8s appr\u00e9ci\u00e9e pour sa capacit\u00e9 \u00e0 <\/span><b>d\u00e9tecter des anomalies de mani\u00e8re non supervis\u00e9e<\/b><span style=\"font-weight: 400;\"> tout en obtenant des r\u00e9sultats satisfaisants dans de nombreux domaines. On peut y trouver des applications en fraude bancaire, diagnostic m\u00e9dical, ou bien en analyse de d\u00e9faut structurel.<\/span><\/p><p><span style=\"font-weight: 400;\">Si vous voulez en savoir plus, n\u2019h\u00e9sitez pas \u00e0 jeter un \u0153il \u00e0 nos formations de <a href=\"https:\/\/datascientest.com\/formation-data-scientist\">Data Scientist<\/a>, <a href=\"https:\/\/datascientest.com\/formation-data-analyst\">Data Analyst<\/a> et <a href=\"https:\/\/datascientest.com\/formation-data-engineer\">Data Engineer<\/a>.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d486daf elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"d486daf\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formations-data\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">D\u00e9couvrir nos formations sp\u00e9cialis\u00e9s dans la data<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-e137bc2 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e137bc2\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-d089567\" data-id=\"d089567\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-0b47045 elementor-widget elementor-widget-html\" data-id=\"0b47045\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<script type=\"application\/ld+json\">\n{\n  \"@context\": \"https:\/\/schema.org\",\n  \"@type\": \"FAQPage\",\n  \"mainEntity\": [{\n    \"@type\": \"Question\",\n    \"name\": \"Le probl\u00e8me de l\u2019Accuracy Paradox\",\n    \"acceptedAnswer\": {\n      \"@type\": \"Answer\",\n      \"text\": \"Lorsqu\u2019on fait face \u00e0 un probl\u00e8me de classification de classes en Machine Learning, il arrive parfois que le jeu de donn\u00e9es soit\u00a0d\u00e9s\u00e9quilibr\u00e9. Plus pr\u00e9cis\u00e9ment, on constate une proportion bien plus grande d\u2019une classe vis-\u00e0-vis de l\u2019autre au sein de l\u2019\u00e9chantillon d\u2019apprentissage. Ce probl\u00e8me est une caract\u00e9ristique fr\u00e9quente des probl\u00e8mes de classification binaires.\u00a0 On pourrait en premi\u00e8re approche, penser \u00e0 appliquer un algorithme de classification classique, comme les arbres de d\u00e9cisions, les K-NN ou les SVM pr\u00e9sent\u00e9s sur notre blog. Toutefois, ces algorithmes ne sont en r\u00e9alit\u00e9 pas en mesure de g\u00e9rer ces jeux de donn\u00e9es atypiques o\u00f9 les disparit\u00e9s entre les classes sont assez fortes. Ils tendent \u00e0 maximiser par le biais de leur fonction de perte des quantit\u00e9s tel que l\u2019accuracy, mais sans tenir compte de la distribution des donn\u00e9es.\"\n    }\n  },{\n    \"@type\": \"Question\",\n    \"name\": \"L\u2019Isolation Forest : c\u2019est quoi ?\",\n    \"acceptedAnswer\": {\n      \"@type\": \"Answer\",\n      \"text\": \"Mais alors comment faire face \u00e0 ce probl\u00e8me qui vient ruiner les efforts de plusieurs algorithmes classiques rondement entra\u00een\u00e9s ? Il existe plusieurs solutions comme par exemple les m\u00e9thodes de r\u00e9-\u00e9chantillonnages comme l'undersampling ou l'oversampling qui s\u00e9lectionnent al\u00e9atoirement des donn\u00e9es afin d\u2019\u00e9quilibrer le ratio de classe. Toutefois, lorsque ces m\u00e9thodes ne sont pas assez efficaces, il est d\u2019usage de repenser le probl\u00e8me et de consid\u00e9rer la d\u00e9tection d\u2019anomalies comme un probl\u00e8me de classification non supervis\u00e9e. C\u2019est l\u00e0 qu\u2019intervient le tant attendu Isolation Forest !\"\n    }\n  }]\n}\n<\/script>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Aujourd\u2019hui, nous nous int\u00e9ressons \u00e0 l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires tels que la d\u00e9tection de fraude ou le diagnostic d\u2019une maladie. Cette technique pr\u00e9sent\u00e9e \u00e0 l\u2019Eighth IEEE International Conference en 2008 est la premi\u00e8re technique de classification d\u00e9di\u00e9e \u00e0 la d\u00e9tection d\u2019anomalies bas\u00e9e sur l\u2019isolation. [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":98755,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2497],"tags":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?<\/title>\n<meta name=\"description\" content=\"l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires comme le diagnostic d&#039;une maladie\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/isolation-forest\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/isolation-forest\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2021-09-10T09:31:25+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-01-25T11:54:40+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/09\/illu_iso_forest-50.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"562\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"J\u00e9r\u00e9my Robert\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"J\u00e9r\u00e9my Robert\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/isolation-forest\",\"url\":\"https:\/\/datascientest.com\/isolation-forest\",\"name\":\"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2021-09-10T09:31:25+00:00\",\"dateModified\":\"2024-01-25T11:54:40+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires comme le diagnostic d'une maladie\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/isolation-forest\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\",\"name\":\"J\u00e9r\u00e9my Robert\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"caption\":\"J\u00e9r\u00e9my Robert\"},\"description\":\"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/jeremyrobert1\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/jeremyrobert\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?","description":"l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires comme le diagnostic d'une maladie","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/datascientest.com\/isolation-forest","og_locale":"fr_FR","og_type":"article","og_title":"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?","og_url":"https:\/\/datascientest.com\/isolation-forest","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2021-09-10T09:31:25+00:00","article_modified_time":"2024-01-25T11:54:40+00:00","og_image":[{"width":1024,"height":562,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/09\/illu_iso_forest-50.png","type":"image\/png"}],"author":"J\u00e9r\u00e9my Robert","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"J\u00e9r\u00e9my Robert","Dur\u00e9e de lecture estim\u00e9e":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/isolation-forest","url":"https:\/\/datascientest.com\/isolation-forest","name":"Isolation Forest : Comment d\u00e9tecter les anomalies dans une dataset ?","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2021-09-10T09:31:25+00:00","dateModified":"2024-01-25T11:54:40+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"l\u2019Isolation Forest, un algorithme de Machine Learning destin\u00e9 \u00e0 r\u00e9soudre des probl\u00e8mes de classification binaires comme le diagnostic d'une maladie","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/isolation-forest"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925","name":"J\u00e9r\u00e9my Robert","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","caption":"J\u00e9r\u00e9my Robert"},"description":"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.","sameAs":["https:\/\/www.linkedin.com\/in\/jeremyrobert1\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/jeremyrobert"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/87971"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=87971"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/87971\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/98755"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=87971"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=87971"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=87971"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}