{"cells": [{"cell_type": "markdown", "metadata": {}, "source": ["# 2A.ml - Imbalanced dataset\n", "\n", "Un jeu de donn\u00e9es *imbalanced* signifie qu'une classe est sous repr\u00e9sent\u00e9e dans un probl\u00e8me de classification. Lire [8 Tactics to Combat Imbalanced Classes in Your Machine Learning Dataset](http://machinelearningmastery.com/tactics-to-combat-imbalanced-classes-in-your-machine-learning-dataset/)."]}, {"cell_type": "code", "execution_count": 1, "metadata": {}, "outputs": [{"data": {"text/html": ["
\n", ""], "text/plain": [""]}, "execution_count": 2, "metadata": {}, "output_type": "execute_result"}], "source": ["from jyquickhelper import add_notebook_menu\n", "add_notebook_menu()"]}, {"cell_type": "code", "execution_count": 2, "metadata": {}, "outputs": [], "source": ["%matplotlib inline"]}, {"cell_type": "markdown", "metadata": {}, "source": ["## G\u00e9n\u00e9ration de donn\u00e9es\n", "\n", "On g\u00e9n\u00e8re un probl\u00e8me de classification binaire avec une classe sous repr\u00e9sent\u00e9e."]}, {"cell_type": "code", "execution_count": 3, "metadata": {}, "outputs": [{"data": {"text/html": ["\n", "\n", "
\n", " \n", " \n", " \n", " X1 \n", " X2 \n", " decision \n", " cl \n", " \n", " \n", " \n", " \n", " count \n", " 1000.000000 \n", " 1000.000000 \n", " 1000.000000 \n", " 1000.000000 \n", " \n", " \n", " mean \n", " 0.501011 \n", " 0.497330 \n", " 1.047568 \n", " 0.080000 \n", " \n", " \n", " std \n", " 0.281113 \n", " 0.295740 \n", " 0.410862 \n", " 0.271429 \n", " \n", " \n", " min \n", " 0.001387 \n", " 0.001413 \n", " 0.032132 \n", " 0.000000 \n", " \n", " \n", " 25% \n", " 0.274936 \n", " 0.238735 \n", " 0.776950 \n", " 0.000000 \n", " \n", " \n", " 50% \n", " 0.505160 \n", " 0.491212 \n", " 1.025373 \n", " 0.000000 \n", " \n", " \n", " 75% \n", " 0.729076 \n", " 0.752400 \n", " 1.340745 \n", " 0.000000 \n", " \n", " \n", " max \n", " 0.999988 \n", " 0.999728 \n", " 2.048328 \n", " 1.000000 \n", " \n", " \n", "
\n", "
"], "text/plain": [" X1 X2 decision cl\n", "count 1000.000000 1000.000000 1000.000000 1000.000000\n", "mean 0.501011 0.497330 1.047568 0.080000\n", "std 0.281113 0.295740 0.410862 0.271429\n", "min 0.001387 0.001413 0.032132 0.000000\n", "25% 0.274936 0.238735 0.776950 0.000000\n", "50% 0.505160 0.491212 1.025373 0.000000\n", "75% 0.729076 0.752400 1.340745 0.000000\n", "max 0.999988 0.999728 2.048328 1.000000"]}, "execution_count": 4, "metadata": {}, "output_type": "execute_result"}], "source": ["import numpy.random\n", "import pandas\n", "\n", "def generate_data(nb, ratio, noise):\n", " mat = numpy.random.random((nb,2))\n", " noise = numpy.random.random((mat.shape[0],1)) * noise\n", " data = pandas.DataFrame(mat, columns=[\"X1\", \"X2\"])\n", " data[\"decision\"] = data.X1 + data.X2 + noise.ravel()\n", " vec = list(sorted(data[\"decision\"]))\n", " l = len(vec)- 1 - int(len(vec) * ratio)\n", " seuil = vec[l]\n", " data[\"cl\"] = data[\"decision\"].apply(lambda r: 1 if r > seuil else 0)\n", " from sklearn.utils import shuffle\n", " data = shuffle(data)\n", " return data\n", "\n", "data = generate_data(1000, 0.08, 0.1)\n", "data.describe()"]}, {"cell_type": "code", "execution_count": 4, "metadata": {}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["ax = data[data.cl==1].plot(x=\"X1\", y=\"X2\", kind=\"scatter\", label=\"cl1\", color=\"r\")\n", "data[data.cl==0].plot(x=\"X1\", y=\"X2\", kind=\"scatter\", label=\"cl0\", ax=ax)\n", "ax.set_title(\"Random imbalanced data\");"]}, {"cell_type": "code", "execution_count": 5, "metadata": {}, "outputs": [], "source": ["from sklearn.model_selection import train_test_split\n", "while True:\n", " X_train, X_test, y_train, y_test = train_test_split(data[[\"X1\", \"X2\"]], data[\"cl\"])\n", " if sum(y_test) > 0:\n", " break"]}, {"cell_type": "markdown", "metadata": {}, "source": ["Le d\u00e9coupage apprentissage est d\u00e9licat car il n'y pas beaucoup d'exemples pour la classe sous-repr\u00e9sent\u00e9e."]}, {"cell_type": "code", "execution_count": 6, "metadata": {}, "outputs": [{"data": {"text/plain": ["22"]}, "execution_count": 7, "metadata": {}, "output_type": "execute_result"}], "source": ["y_test.sum()"]}, {"cell_type": "markdown", "metadata": {}, "source": ["## Apprendre et tester un mod\u00e8le\n", "\n", "Pour ce type de probl\u00e8me, un mod\u00e8le qui retourne la classe majoritaire quelque soit le cas est d\u00e9j\u00e0 un bon mod\u00e8le puisqu'il retourne la bonne r\u00e9ponse dans la majorit\u00e9 des cas."]}, {"cell_type": "code", "execution_count": 7, "metadata": {}, "outputs": [{"data": {"text/html": ["\n", "\n", "
\n", " \n", " \n", " \n", " train:y=0 \n", " train:y=1 \n", " test:y=0 \n", " test:y=1 \n", " \n", " \n", " \n", " \n", " y=0 \n", " 692 \n", " 0 \n", " 228 \n", " 0 \n", " \n", " \n", " y=1 \n", " 34 \n", " 24 \n", " 12 \n", " 10 \n", " \n", " \n", "
\n", "
"], "text/plain": [" train:y=0 train:y=1 test:y=0 test:y=1\n", "y=0 692 0 228 0\n", "y=1 34 24 12 10"]}, "execution_count": 8, "metadata": {}, "output_type": "execute_result"}], "source": ["from sklearn.metrics import confusion_matrix\n", "\n", "def confusion(model, X_train, X_test, y_train, y_test):\n", " model.fit(X_train, y_train)\n", " predt = model.predict(X_train)\n", " c_train = confusion_matrix(y_train, predt)\n", " pred = model.predict(X_test)\n", " c_test = confusion_matrix(y_test, pred)\n", " return pandas.DataFrame(numpy.hstack([c_train, c_test]), index=[\"y=0\", \"y=1\"],\n", " columns=\"train:y=0 train:y=1 test:y=0 test:y=1\".split())\n", "\n", "from sklearn.linear_model import LogisticRegression\n", "confusion(LogisticRegression(solver='lbfgs'),\n", " X_train, X_test, y_train, y_test)"]}, {"cell_type": "markdown", "metadata": {}, "source": ["Quelques exemples pour tester, quelques exemples pour apprendre. C'est peu."]}, {"cell_type": "code", "execution_count": 8, "metadata": {}, "outputs": [{"data": {"text/html": ["\n", "\n", "
\n", " \n", " \n", " \n", " train:y=0 \n", " train:y=1 \n", " test:y=0 \n", " test:y=1 \n", " \n", " \n", " \n", " \n", " y=0 \n", " 692 \n", " 0 \n", " 227 \n", " 1 \n", " \n", " \n", " y=1 \n", " 0 \n", " 58 \n", " 2 \n", " 20 \n", " \n", " \n", "
\n", "
"], "text/plain": [" train:y=0 train:y=1 test:y=0 test:y=1\n", "y=0 692 0 227 1\n", "y=1 0 58 2 20"]}, "execution_count": 9, "metadata": {}, "output_type": "execute_result"}], "source": ["from sklearn.tree import DecisionTreeClassifier\n", "confusion(DecisionTreeClassifier(), X_train, X_test, y_train, y_test)"]}, {"cell_type": "code", "execution_count": 9, "metadata": {}, "outputs": [{"data": {"text/html": ["\n", "\n", "
\n", " \n", " \n", " \n", " train:y=0 \n", " train:y=1 \n", " test:y=0 \n", " test:y=1 \n", " \n", " \n", " \n", " \n", " y=0 \n", " 692 \n", " 0 \n", " 227 \n", " 1 \n", " \n", " \n", " y=1 \n", " 1 \n", " 57 \n", " 2 \n", " 20 \n", " \n", " \n", "
\n", "
"], "text/plain": [" train:y=0 train:y=1 test:y=0 test:y=1\n", "y=0 692 0 227 1\n", "y=1 1 57 2 20"]}, "execution_count": 10, "metadata": {}, "output_type": "execute_result"}], "source": ["from sklearn.ensemble import RandomForestClassifier\n", "confusion(RandomForestClassifier(n_estimators=10),\n", " X_train, X_test, y_train, y_test)"]}, {"cell_type": "markdown", "metadata": {}, "source": ["L'algorithme de la r\u00e9gression logistique converge plus difficile que celui des arbres de d\u00e9cision. Voyons comment cela \u00e9volue entre de la norme *L1* ou *L2* et de la proportion de la classe mal balanc\u00e9e."]}, {"cell_type": "code", "execution_count": 10, "metadata": {}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["ratio = list(_/400.0 for _ in range(1, 101))\n", "rows = []\n", "for r in ratio:\n", " data = generate_data(1000, r, noise=0.0)\n", " while True:\n", " X_train, X_test, y_train, y_test = train_test_split(data[[\"X1\", \"X2\"]], data[\"cl\"])\n", " if sum(y_test) > 0 and sum(y_train) > 0:\n", " break\n", " c = confusion(LogisticRegression(penalty='l1', solver='liblinear'),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row = dict(ratio=r, precision_l1=c1 / (c0 + c1) )\n", " c = confusion(LogisticRegression(penalty='l2', solver=\"liblinear\"),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row[\"precision_l2\"] = c1 / (c0 + c1)\n", " rows.append(row)\n", "df = pandas.DataFrame(rows)\n", "\n", "import matplotlib.pyplot as plt\n", "fig, ax = plt.subplots(1, 1)\n", "df.plot(x=\"ratio\", y=[_ for _ in df.columns if _ !=\"ratio\"], ax=ax)\n", "ax.set_xlabel(\"Ratio classe mal balanc\u00e9e\")\n", "ax.set_ylabel(\"Pr\u00e9cision pour la classe 1 (petite classe)\");"]}, {"cell_type": "markdown", "metadata": {}, "source": ["La norme **l1** est plus sensible aux petites classes. La m\u00e9trique [balanced_accuracy_score](https://scikit-learn.org/stable/modules/generated/sklearn.metrics.balanced_accuracy_score.html) calcule la performance du mod\u00e8le en donnant le m\u00eame poids quelque soit la taille de la classe, il fait la moyenne de l'*accuracy* par classe."]}, {"cell_type": "code", "execution_count": 11, "metadata": {}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["from sklearn.metrics import balanced_accuracy_score, accuracy_score\n", "\n", "ratio = list(_/400.0 for _ in range(1, 101))\n", "rows = []\n", "for r in ratio:\n", " data = generate_data(1000, r, noise=0.0)\n", " while True:\n", " X_train, X_test, y_train, y_test = train_test_split(data[[\"X1\", \"X2\"]], data[\"cl\"])\n", " if sum(y_test) > 0 and sum(y_train) > 0:\n", " break\n", " model = LogisticRegression(penalty='l2', solver='liblinear')\n", " model.fit(X_train, y_train)\n", " predt = model.predict(X_test)\n", " bacc_l2 = balanced_accuracy_score(y_test, predt)\n", " acc_l2 = accuracy_score(y_test, predt)\n", " \n", " model = LogisticRegression(penalty='l1', solver='liblinear')\n", " model.fit(X_train, y_train)\n", " predt = model.predict(X_test)\n", " bacc_l1 = balanced_accuracy_score(y_test, predt)\n", " acc_l1 = accuracy_score(y_test, predt)\n", "\n", " row = dict(standard_l2=acc_l2, balanced_l2=bacc_l2, ratio=r,\n", " standard_l1=acc_l1, balanced_l1=bacc_l1)\n", " rows.append(row)\n", "df = pandas.DataFrame(rows)\n", "\n", "fig, ax = plt.subplots(1, 1)\n", "df.plot(x=\"ratio\", y=[_ for _ in df.columns if _ !=\"ratio\"], ax=ax)\n", "ax.set_xlabel(\"Ratio classe mal balanc\u00e9e\")\n", "ax.set_ylabel(\"Accuracy\");"]}, {"cell_type": "markdown", "metadata": {}, "source": ["La m\u00e9trique classique \"accuracy\" ne permet pas de d\u00e9tecter un probl\u00e8me de classification lorsqu'une classe est mal balanc\u00e9e car chaque exemple est pond\u00e9r\u00e9 de la m\u00eame fa\u00e7on. L'*accuracy* est donc tr\u00e8s proche de celle obtenue sur la classe majoritaire."]}, {"cell_type": "code", "execution_count": 12, "metadata": {"scrolled": true}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["from sklearn.ensemble import AdaBoostClassifier\n", "ratio = list(_/400.0 for _ in range(1, 101))\n", "rows = []\n", "for r in ratio:\n", " data = generate_data(1000, r, noise=0.0)\n", " while True:\n", " X_train, X_test, y_train, y_test = train_test_split(data[[\"X1\", \"X2\"]], data[\"cl\"])\n", " if sum(y_test) > 0 and sum(y_train) > 0:\n", " break\n", " c = confusion(LogisticRegression(penalty='l1', solver=\"liblinear\"),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row = dict(ratio=r, precision_l1=c1 / (c0 + c1) )\n", " c = confusion(LogisticRegression(penalty='l2', solver=\"liblinear\"),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row[\"precision_l2\"] = c1 / (c0 + c1)\n", " c = confusion(AdaBoostClassifier(LogisticRegression(penalty='l2', solver=\"liblinear\"),\n", " algorithm=\"SAMME.R\", n_estimators=50,\n", " learning_rate=3),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row[\"pre_AdaBoost_l2-50\"] = c1 / (c0 + c1)\n", " c = confusion(AdaBoostClassifier(LogisticRegression(penalty='l2', solver=\"liblinear\"),\n", " algorithm=\"SAMME.R\", n_estimators=100,\n", " learning_rate=3),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row[\"prec_AdaBoost_l2-100\"] = c1 / (c0 + c1)\n", " rows.append(row)\n", "df = pandas.DataFrame(rows)\n", "\n", "fig, ax = plt.subplots(1, 1)\n", "df.plot(x=\"ratio\", y=[_ for _ in df.columns if _ != \"ratio\"], ax=ax)\n", "ax.set_xlabel(\"Ratio classe mal balanc\u00e9e\")\n", "ax.set_ylabel(\"Pr\u00e9cision pour la classe 1 (petite classe)\");"]}, {"cell_type": "markdown", "metadata": {}, "source": ["On voit que l'algorithme [AdaBoost](https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostClassifier.html#sklearn.ensemble.AdaBoostClassifier) permet de favoriser les petites classes mais il faut jouer avec le learning rate et le nombre d'estimateurs."]}, {"cell_type": "code", "execution_count": 13, "metadata": {}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["from sklearn.ensemble import AdaBoostClassifier\n", "ratio = list(_/400.0 for _ in range(1, 101))\n", "rows = []\n", "for r in ratio:\n", " data = generate_data(1000, r, noise=0.0)\n", " while True:\n", " X_train, X_test, y_train, y_test = train_test_split(data[[\"X1\", \"X2\"]], data[\"cl\"])\n", " if sum(y_test) > 0 and sum(y_train) > 0:\n", " break\n", " c = confusion(DecisionTreeClassifier(),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row = dict(ratio=r, precision_tree=c1 / (c0 + c1) )\n", " c = confusion(RandomForestClassifier(n_estimators=10),\n", " X_train, X_test, y_train, y_test)\n", " c0, c1 = c.loc[\"y=1\", \"test:y=0\"], c.loc[\"y=1\", \"test:y=1\"]\n", " row[\"precision_rf\"] = c1 / (c0 + c1)\n", " rows.append(row)\n", " \n", "fig, ax = plt.subplots(1, 1)\n", "df = pandas.DataFrame(rows)\n", "df.plot(x=\"ratio\", y=[_ for _ in df.columns if _ != \"ratio\"], ax=ax)\n", "ax.set_xlabel(\"Ratio classe mal balanc\u00e9e\")\n", "ax.set_ylabel(\"Pr\u00e9cision pour la classe 1 (petite classe)\");"]}, {"cell_type": "markdown", "metadata": {}, "source": ["Les m\u00e9thodes ensemblistes fonctionnent mieux dans ce cas car l'algorithme cherche la meilleure s\u00e9paration entre deux classes de fa\u00e7on \u00e0 ce que les deux classes soient de chaque c\u00f4t\u00e9 de cette fronti\u00e8re. La proportion d'exemples a moins d'importance pour le crit\u00e8re de [Gini](https://fr.wikipedia.org/wiki/Coefficient_de_Gini). Dans l'exemple suivant, on trie selon la variable $X_1$ et on cherche la meilleur s\u00e9paration"]}, {"cell_type": "code", "execution_count": 14, "metadata": {}, "outputs": [{"data": {"text/plain": ["array([[0.0133622 , 0.00996042, 0.04423642, 0. ],\n", " [0.0215605 , 0.0276332 , 0.21230519, 0. ],\n", " [0.02827806, 0.03093092, 0.2737886 , 0. ],\n", " [0.03648334, 0.03178925, 0.31218421, 0. ],\n", " [0.03649288, 0.03712127, 0.34506943, 0. ]])"]}, "execution_count": 15, "metadata": {}, "output_type": "execute_result"}], "source": ["data = generate_data(100, 0.08, 0.1).values\n", "data.sort(axis=0)\n", "data[:5]"]}, {"cell_type": "code", "execution_count": 15, "metadata": {}, "outputs": [{"data": {"text/plain": ["array([5.20836364e+01, 9.52380952e-03, 4.28571429e-02, 1.26190476e-01,\n", " 4.28571429e-02, 2.61904762e-02, 2.55640000e+02])"]}, "execution_count": 16, "metadata": {}, "output_type": "execute_result"}], "source": ["from ensae_teaching_cs.ml.gini import gini\n", "\n", "def gini_gain_curve(Y):\n", " \"le code n'est pas le plus efficace du monde mais \u00e7a suffira\"\n", " g = gini(Y)\n", " curve = numpy.empty((len(Y),))\n", " for i in range(1, len(Y)-1):\n", " g1 = gini(Y[:i])\n", " g2 = gini(Y[i:])\n", " curve[i] = g - (g1 + g2) / 2\n", " return curve\n", "\n", "gini_gain_curve([0, 1, 0, 1, 1, 1, 1])"]}, {"cell_type": "code", "execution_count": 16, "metadata": {}, "outputs": [{"data": {"text/plain": ["0.9550000000000001"]}, "execution_count": 17, "metadata": {}, "output_type": "execute_result"}], "source": ["from ensae_teaching_cs.ml.gini import gini\n", "gini(data[:, 3])"]}, {"cell_type": "code", "execution_count": 17, "metadata": {}, "outputs": [{"data": {"image/png": "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\n", "text/plain": [""]}, "metadata": {"needs_background": "light"}, "output_type": "display_data"}], "source": ["fig, ax = plt.subplots(1, 1)\n", "for skew in [0.05, 0.1, 0.15, 0.2, 0.25, 0.5]:\n", " data = generate_data(100, skew, 0.1).values\n", " data.sort(axis=0)\n", " ax.plot(gini_gain_curve(data[:, 3]), label=\"balance=%f\" % skew)\n", "ax.legend()\n", "ax.set_title(\"Gini gain for different class proportion\")\n", "ax.set_ylabel(\"Gini\")\n", "ax.set_xlabel(\"x\");"]}, {"cell_type": "markdown", "metadata": {}, "source": ["Ce n'est pas vraiment pas l'algorithme des arbres de d\u00e9cision mais l'id\u00e9e est de montrer que les arbres de d\u00e9cision sont moins sensibles aux petites classes quand il s'agit de trouver la meilleure s\u00e9paration. Et c'est n\u00e9cessaire car pour les branches les plus basses, tous les sous-\u00e9chantillons qui terminent dans ces branches sont tr\u00e8s mal balanc\u00e9s."]}, {"cell_type": "markdown", "metadata": {}, "source": ["## Exercice 1 : r\u00e9duire les exemples loin des fronti\u00e8res\n", "\n", "Pour r\u00e9\u00e9quilibrer la proportion des classes, on cherche \u00e0 enlever des points de la base d'apprentissage pour lesquels il n'y a pas d'ambigu\u00eft\u00e9, c'est-\u00e0-dire loin des fronti\u00e8res. Imaginer une solution \u00e0 l'aides des [k plus proches voisins](http://scikit-learn.org/stable/modules/generated/sklearn.neighbors.KNeighborsClassifier.html)."]}, {"cell_type": "code", "execution_count": 18, "metadata": {}, "outputs": [], "source": []}, {"cell_type": "markdown", "metadata": {}, "source": ["## Exercice 2 : multiplier les exemples\n", "\n", "L'id\u00e9e est d'utiliser une technique pour multiplier les exemples de la classe sous-repr\u00e9sent\u00e9e sans pour autant avoir des exemples exactement identiques. On utilise l'algorithme [SMOTE](http://jair.org/media/953/live-953-2037-jair.pdf). En r\u00e9sum\u00e9, l'algorithme consiste \u00e0 cr\u00e9er des exemples pour la classe sous-repr\u00e9sent\u00e9e. On choisit un de ces exemples $X$. Pour cet $X$, on calcule ses $k$ plus proches voisins dans la base d'apprentissage, toutes classes comprises. On choisit un voisin al\u00e9atoire $V$ parmi les $k$ voisins. On tire un nombre al\u00e9aloire $h\\in]0,1]$. Le nouvel \u00e9l\u00e9ment ajout\u00e9 \u00e0 la base d'apprentissage est $X + h (V-X)$ et il est associ\u00e9 \u00e0 la classe sous-repr\u00e9sent\u00e9e. On continue jusqu'\u00e0 la proportion souhait\u00e9e.\n", "\n", "* [SMOTE: Synthetic Minority Over-sampling Technique](http://jair.org/media/953/live-953-2037-jair.pdf)\n", "* [Detecting Click Fraud in Online Advertising: A Data Mining Approach](http://www.jmlr.org/papers/volume15/oentaryo14a/oentaryo14a.pdf)\n", "* [On the Statistical Consistency of Algorithms for Binary Classification under Class Imbalance](http://jmlr.org/proceedings/papers/v28/menon13a.pdf)\n", "* [On multi-class classification through the minimization of the confusion matrix norm](http://jmlr.org/proceedings/papers/v29/Koco13.pdf)"]}, {"cell_type": "code", "execution_count": 19, "metadata": {}, "outputs": [], "source": []}, {"cell_type": "markdown", "metadata": {"collapsed": true}, "source": ["## Exercice 3 : essai du module imbalanced\n", "\n", "Ce module impl\u00e9mente diff\u00e9rentes fa\u00e7ons de g\u00e9rer les classes sous et sur-repr\u00e9sent\u00e9es.\n", "\n", "* [algorithmes impl\u00e9ment\u00e9s](https://github.com/scikit-learn-contrib/imbalanced-learn#about)\n", "* [exemples - imbalanced-learn](http://contrib.scikit-learn.org/imbalanced-learn/stable/auto_examples/index.html)"]}, {"cell_type": "code", "execution_count": 20, "metadata": {}, "outputs": [], "source": []}, {"cell_type": "markdown", "metadata": {}, "source": ["## Exercice 4 : validation crois\u00e9e\n", "\n", "Lorsqu'une classe est sous repr\u00e9sent\u00e9e, la validation crois\u00e9e doit \u00eatre effectu\u00e9e sous contrainte. Si elle est r\u00e9alis\u00e9e de fa\u00e7on compl\u00e8tement al\u00e9atoire, il est probable que la classe sous repr\u00e9sent\u00e9e ne soit pas pr\u00e9sente. Si la classe 0 poss\u00e8de $k$ exemples parmi $N$, quelle est la distribution du minimum d'observations dans une des clasees ? Il veut comparer une crossvalidation classique avec un \u00e9chantillon stratigi\u00e9e ([StratifiedKFold](http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.StratifiedKFold.html))."]}, {"cell_type": "code", "execution_count": 21, "metadata": {}, "outputs": [], "source": []}, {"cell_type": "code", "execution_count": 22, "metadata": {}, "outputs": [], "source": []}], "metadata": {"kernelspec": {"display_name": "Python 3", "language": "python", "name": "python3"}, "language_info": {"codemirror_mode": {"name": "ipython", "version": 3}, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.7.2"}}, "nbformat": 4, "nbformat_minor": 2}