\n",
"In diesem Anwendungsbeispiel vergleichen wir Ridge Regression mit einer C-SVM. Die Daten wurden künstlich erstellt und dienen nur als Veranschaulichung einer potentiellen echten Anwendung.\n",
"
\n",
"Ein neues Medikament kommt auf den Markt, welches sich gut für die Therapie eines speziellen Tumors eignet. Jedoch hat man festgestellt, dass es bei einer Gruppe von Patienten schneller wirkt als bei einer anderen. Leider kommt es bei einer längeren Behandlung mit dem Medikament zu erheblichen Nebenwirkungen. Daher möchte man vermeiden, dass Patienten das Medikament verabreicht wird, wenn es eine zu lange Reaktionszeit benötigen würde. \n",
"
\n",
" \n",
"
\n",
"Es wird vermutet, dass verschiedene genetische Veränderungen (Mutationen) im Genom der einzelnen Patienten eine Rolle spielen könnten, weshalb das Medikament eine schnellere oder längere Reaktionszeit hat. \n",
"
\n",
" \n",
"
\n",
"Unsere Aufgabe als Machine Learner ist es nun, ein Vorhersagemodell zu entwickeln, welches uns anhand der genetischen Unterschiede zwischen den Patienten die Reaktionszeit (Anzahl der Tage) vorhersagt, ab wann das Medikament für den individuellen Patienten zu wirken beginnt. \n",
"
\n",
" \n",
"
\n",
"Um die Reaktionszeit in Tagen vorherzusagen, bekommen wir von unserem Auftragssteller einen Datensatz von 400 Patienten. Für jeden Patienten wurden 600 verschiedene genetische Veränderungen (Mutationen) an derselben Stelle im Genom gemessen. Zusätzlich wurde die Anzahl der Tage notiert, ab wann das Medikament für den individuellen Patient begonnen hat zu wirken.\n",
"
\n",
"\n",
"\n",
"## 1. Vorhersage der genauen Reaktionszeit mit Ridge Regression\n",
"
\n",
"Um die Reaktionszeit des Medikaments für neue Patienten vorherzusagen, verwenden wir Ridge Regression, da es sich um ein Regressionproblem handelt und wir mehr Features als Patienten haben. Zudem verwenden wir Ridge Regression, um ein Modell zu lernen, welches nicht überangepasst ist (overfitted) und gut auf unbekannte Patienten generalisiert, indem wir den optimalen Regularisierungsparameter lernen.\n",
"
\n",
"### 1.1 Daten Vorverarbeitung"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"scrolled": true
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Orginal Daten\n",
"Anzahl Patienten:\t400\n",
"Anzahl Features:\t600\n",
"\n",
"Trainingsdaten\n",
"Anzahl Patienten:\t320\n",
"Anzahl Features:\t600\n",
"\n",
"Testdaten\n",
"Anzahl Patienten:\t80\n",
"Anzahl Features:\t600\n"
]
}
],
"source": [
"%matplotlib inline\n",
"import scipy as sp\n",
"import matplotlib\n",
"import pylab as pl\n",
"matplotlib.rcParams.update({'font.size': 15})\n",
"\n",
"from sklearn.linear_model import Ridge\n",
"from sklearn.svm import SVC\n",
"from sklearn.model_selection import KFold, StratifiedKFold, GridSearchCV,StratifiedShuffleSplit\n",
"from sklearn.model_selection import cross_val_score, train_test_split\n",
"from sklearn.metrics import accuracy_score, mean_squared_error, mean_absolute_error\n",
"\n",
"random_state = 42\n",
"\n",
"#Lade Simulierte Daten\n",
"data = sp.loadtxt(\"toydata/X.txt\")\n",
"binary_target = sp.loadtxt(\"toydata/y_binary.txt\")\n",
"continuous_target = sp.loadtxt(\"toydata/y.txt\")\n",
"\n",
"#Zusammenfassung der Daten\n",
"print(\"Orginal Daten\")\n",
"print(\"Anzahl Patienten:\\t%d\"%data.shape[0])\n",
"print(\"Anzahl Features:\\t%d\"%data.shape[1])\n",
"print\n",
"\n",
"#Splitte Daten in Trainings und Test Daten\n",
"train_test_data = train_test_split(data,\n",
" continuous_target,\n",
" test_size=0.2,\n",
" random_state=random_state)\n",
"training_data = train_test_data[0]\n",
"testing_data = train_test_data[1]\n",
"training_target = train_test_data[2]\n",
"testing_target = train_test_data[3]\n",
"\n",
"print(\"Trainingsdaten\")\n",
"print(\"Anzahl Patienten:\\t%d\"%training_data.shape[0])\n",
"print(\"Anzahl Features:\\t%d\"%training_data.shape[1])\n",
"print\n",
"print(\"Testdaten\")\n",
"print(\"Anzahl Patienten:\\t%d\"%testing_data.shape[0])\n",
"print(\"Anzahl Features:\\t%d\"%testing_data.shape[1])"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 1.2 Trainiere Ridge Regression auf Trainingsdaten\n",
"\n",
"Als Erstes trainieren wir ein Ridge Regression Modell auf den Trainingsdaten mit einer 5-fachen Kreuzvalidierung und einer interenen Liniensuche, um das optimiale $\\alpha$ zu finden. Zusätzlich plotten wir den Trainingsfehler gegen den Subtestfehler, um den Effekt von unterschiedlichen $\\alpha$'s zu betrachten."
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"5-fache Kreuzvalidierung mit interner Liniensuche auf Trainingsdaten\n",
"Durchschnitt(Mean Squared Error):\t\t587.09 (-+ 53.45)\n",
"\n"
]
},
{
"data": {
"image/png": "iVBORw0KGgoAAAANSUhEUgAABD4AAAHJCAYAAABt+PnTAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"#Initialisiere Alphas für Ridge Regression\n",
"alphas = sp.logspace(-2,8,11)\n",
"param_grid = dict(alpha=alphas)\n",
"\n",
"#5-fach Kreuzvalidierung um den mittleren quadratischen Fehler zu berechnen\n",
"outer_cv = KFold(n_splits=5,shuffle=True,random_state=random_state)\n",
"\n",
"#Liniensuche um das optimale Alpha zu finden\n",
"line_search = GridSearchCV(Ridge(random_state=random_state,solver=\"cholesky\"),\n",
" param_grid=param_grid,\n",
" scoring=\"neg_mean_squared_error\")\n",
"\n",
"#Führe 5-fach Kreuzvalidierung mit interner Liniensuche aus und berechne mittleren quadratischen Fehler\n",
"score = cross_val_score(line_search,X=training_data,y=training_target,cv=outer_cv,scoring=\"neg_mean_squared_error\")\n",
"print(\"5-fache Kreuzvalidierung mit interner Liniensuche auf Trainingsdaten\")\n",
"print(\"Durchschnitt(Mean Squared Error):\\t\\t%.2f (-+ %.2f)\"%(score.mean()*(-1),score.std()))\n",
"print\n",
"\n",
"#Berechne optimales Alpha auf allen Trainings Daten\n",
"line_search.fit(training_data,training_target)\n",
"optimal_alpha = line_search.best_params_['alpha']\n",
"\n",
"#Visualisiere Trainings und Validierungs Fehler für unterschiedliche Alphas\n",
"pl.figure(figsize=(18,7))\n",
"pl.plot(alphas,line_search.cv_results_['mean_train_score']*(-1),label=\"Trainingsfehler\",color=\"#e67e22\")\n",
"pl.fill_between(alphas,\n",
" line_search.cv_results_['mean_train_score']*(-1)-line_search.cv_results_['std_train_score'],\n",
" line_search.cv_results_['mean_train_score']*(-1)+line_search.cv_results_['std_train_score'],\n",
" alpha=0.3,color=\"#e67e22\")\n",
"pl.plot(alphas,line_search.cv_results_['mean_test_score']*(-1),label=\"Subtestfehler\",color=\"#2980b9\")\n",
"pl.fill_between(alphas,\n",
" line_search.cv_results_['mean_test_score']*(-1)-line_search.cv_results_['std_test_score'],\n",
" line_search.cv_results_['mean_test_score']*(-1)+line_search.cv_results_['std_test_score'],\n",
" alpha=0.3,color=\"#2980b9\")\n",
"#pl.xlim(0,max(alphas))\n",
"pl.xscale(\"log\")\n",
"pl.ylabel(\"Mean Squared Error\")\n",
"pl.xlabel(\"Alphas\")\n",
"pl.legend(frameon=True)\n",
"pl.grid(True)\n",
"pl.axvline(x=optimal_alpha,color='r',linestyle=\"--\")\n",
"pl.title(\"Training- vs. Subtest-Fehler (Optimales Alpha = %.1e)\"%optimal_alpha);"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### 1.3 Trainiere Ridge Regression mit optimalem $\\alpha$ und teste auf Testdaten\n",
"Als Nächstes trainieren wir auf den gesamten Trainingsdaten ein Ridge Regression Modell mit dem optimalen $\\alpha$, welches wir in Abschnitt 1.2 mithilfe einer 5-fachen Kreuzvalidierung und einer interenen Liniensuche gefunden haben. Danach testen wir unser Modell auf den nicht verwendeten Testdaten, um zu sehen, wie gut unser Modell generalisiert."
]
},
{
"cell_type": "code",
"execution_count": 3,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Ergebnisse mit optimalem Alpha auf Testdaten\n",
"MSE (Test Daten, Alpha=Optimal):\t699.56 \n",
"Optimales Alpha:\t\t\t100000.00\n",
"\n"
]
}
],
"source": [
"#Trainiere Ridge Regression auf allen Trainings Daten mit optimalen Alpha\n",
"model = Ridge(alpha=optimal_alpha,solver=\"cholesky\")\n",
"model.fit(training_data,training_target)\n",
"\n",
"#Sage auf unbekannten Test Daten vorher\n",
"predictions = model.predict(testing_data)\n",
"print(\"Ergebnisse mit optimalem Alpha auf Testdaten\")\n",
"print(\"MSE (Test Daten, Alpha=Optimal):\\t%.2f \"%(mean_squared_error(testing_target,predictions)))\n",
"print(\"Optimales Alpha:\\t\\t\\t%.2f\"%optimal_alpha)\n",
"print"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
" \n",
"
\n",
"Eine 5-fache Kreuzvalidierung auf den Trainingsdaten führt zu einem Mean Squared Error von $MSE=587.09 \\pm 53.54$. Auf den Testdaten erhalten wir einen Fehler von $MSE=699.56$ ($\\sim 26.5$ Tage). Das bedeutet, dass das Ridge Regression Modell zu eher schlechten Vorhersagen führt (trotz Parameteroptimierung). \n",
"Das kann z. B. daran liegen, dass die vorhandenen Features die genaue Zielvariable (Anzahl der Tage bis das Medikament wirkt) nur unzureichend beschreibt. \n",
"
\n",
" \n",
"
\n",
"Wegen der eher entäuschenden Ergebnisse besprechen sich unsere Auftragsgeber nochmals. Nach längerer Beratung erklären uns diese, dass es ihnen eigentlich nicht auf die genaue Anzahl der Tage ankomme. Vielmehr würde ihnen ausreichen, wenn wir die Patienten vorhersagen könnten, welche eine längere bzw. eine kürzere Reaktionszeit haben. \n",
"Laut den Erfahrungen unseres Auftraggebers ist ab circa 50 Tagen Anwendung mit erheblichen Nebenwirkungen zu rechen. Wir binarisieren unsere Daten, indem alle Patienten mit bis zu 50 Tagen in Klasse 0 und alle Patienten mit mehr als 50 Tagen in Klasse 1 aufgeteilt werden. \n",
"Somit können wir statt einer Regression einen Klassifikationsalgorithmus verwenden, welcher versucht, die Patienten in diese zwei unterschiedlichen Klassen zu klassifizieren.\n",
"