# Importamos las librerias principales import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns sns.set(context='notebook')

titanic = pd.read_csv('https://web.stanford.edu/class/archive/cs/cs109/cs109.1166/stuff/titanic.csv', sep=',')

# Visualizacion del Dataframe titanic.head(10)

# Eliminamos las columnas que no nos interesan titanic.drop( ['Name', 'Fare'], axis=1, inplace=True)

# Renombramos las columnas Siblings/Spouses Aboard y Parents/Children Aboard por SibSp y ParCh titanic.columns = ['Survived', 'Pclass', 'Sex', 'Age', 'SibSp', 'ParCh'] titanic.head()

# Analizamos el shape del objeto titanic.shape

# Visualizamos los tipos de datos titanic.info()

# Cambiamos los tipos de datos de Sex titanic = pd.get_dummies(titanic, columns=['Sex'], drop_first=True)

titanic.dtypes

titanic.head()

# Renombramos columna Sex_male a Sex titanic.rename(columns={'Sex_male':'Sex'},inplace=True)

# Ordenar colummas por nombres titanic = titanic[['Survived','Pclass','Sex','Age','SibSp','ParCh']] titanic

# Vemos la proporcion del a variable objetivo titanic.Survived.value_counts(normalize=True)

# importamos libreria para balancear los datos from imblearn.under_sampling import RandomUnderSampler undersample = RandomUnderSampler(random_state=42)

# Separamos en X e y X_titanic = titanic.drop('Survived', axis=1) y_titanic = titanic.Survived

# Balanceamos los datos X_over_titanic, y_over_titanic = undersample.fit_resample(X_titanic,y_titanic) y_over_titanic.value_counts(normalize=True)

# Importamos las librerias para dividir el dataset from sklearn.model_selection import train_test_split # 30% para test y 70% para train X_train_titanic, X_test_titanic, y_train_titanic, y_test_titanic = train_test_split(X_over_titanic,y_over_titanic, test_size=0.30, random_state=42)

# Importamos las librerias para la creacion del modelo from sklearn.tree import DecisionTreeClassifier from sklearn.model_selection import GridSearchCV # Definir el clasificador y los valores de los hiperparámetros a probar clf = DecisionTreeClassifier(random_state=42) param_grid = {'criterion': ['gini', 'entropy'], 'max_depth': [2, 3, 4, 5]} # Realizar la búsqueda de hiperparámetros utilizando GridSearchCV grid_search = GridSearchCV(clf, param_grid=param_grid, cv=10, return_train_score=True) grid_search.fit(X_train_titanic, y_train_titanic) # Imprimir los resultados print("Mejores hiperparámetros encontrados:") print(grid_search.best_params_) print("Mejor puntuación de validación cruzada:") print(grid_search.best_score_)

# Modelo decision tree con parametros optimizados best_clf = grid_search.best_estimator_

# Predecimos Y y_train_pred_titanic = best_clf.predict(X_train_titanic) y_test_pred_titanic = best_clf.predict(X_test_titanic)

# Graficamos matriz de confusion from sklearn.metrics import confusion_matrix from sklearn.metrics import ConfusionMatrixDisplay cm = confusion_matrix(y_test_titanic,y_test_pred_titanic,labels=best_clf.classes_) ConfusionMatrixDisplay(cm, display_labels=best_clf.classes_).plot()

# Calculo de las predicciones en Train y test y_train_pred = best_clf.predict(X_train_titanic) y_test_pred = best_clf.predict(X_test_titanic)

from sklearn.metrics import accuracy_score print('El accuracy en train es:',accuracy_score(y_train_titanic,y_train_pred_titanic)) print('El accuracy en test es:', accuracy_score(y_test_titanic,y_test_pred_titanic))

feature_scores_titanic = pd.DataFrame(pd.Series(grid_search.best_estimator_.feature_importances_, index=X_train_titanic.columns).sort_values(ascending=False)).T

plt.figure(figsize=(12,4)) sns.barplot(data=feature_scores_titanic) for index, value in enumerate(feature_scores_titanic.values.flatten()): plt.annotate(f'{value:.2f}', xy=(index, value), ha='center', va='bottom') plt.title("Factores clave en la predicción de la supervivencia en el Titanic") plt.show()

import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns sns.set(style='whitegrid', context='notebook')

df_car = pd.read_csv('https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data', header=None) columns_names = ['price','maint','doors','persons','lug_boot','safety','Class'] df_car.columns = columns_names

# Visualizacion del DataFrame df_car.head()

# Visualizacion del shape del DF df_car.shape

# Tipos de datos df_car.info()

# Vereificamos valores missings df_car.isnull().sum()

df_car.drop_duplicates()

df_car.isnull().sum()

# Funcion conteo y proporcion de datos def dist(df,target): count= df[target].value_counts(normalize=False) prop = df[target].value_counts(normalize=True) dist = pd.DataFrame({'Freq[N]':count,'Prop[%]':prop.round(3)}) return dist

# Ver el conteo y la proporción de Class for i in columns_names: print(' '*7,i.upper()) print(dist(df_car,i)) print("*"*23)

# Graficamos la variable Class separada por los distintos atributos fig, axes = plt.subplots(nrows=2, ncols=3, figsize=(12, 8)) for i, variable in enumerate(columns_names[:-1]): row = i % 2 col = i // 2 sns.countplot(data=df_car, x='Class',hue=variable, ax=axes[row][col]) axes[row][col].set_title(f"Evaluation Classes by {variable} Category") plt.tight_layout() plt.show()

from imblearn.under_sampling import RandomUnderSampler

undersample = RandomUnderSampler(random_state=42)

# Separamos datos por X e Y X_car = df_car.drop('Class',axis=1) y_car = df_car.Class

# Se realiza el undersampling X_car, y_car = undersample.fit_resample(X_car,y_car)

import category_encoders as ce encoder = ce.OrdinalEncoder(cols=columns_names[:-1]) X_car = encoder.fit_transform(X_car)

X_car.head()

X_car.dtypes

from sklearn.model_selection import train_test_split X_train_car, X_test_car, y_train_car, y_test_car = train_test_split(X_car,y_car,test_size=0.3, random_state=42)

print('X:',X_train_car.shape, X_test_car.shape) print('y:',y_train_car.shape, y_test_car.shape)

from sklearn.tree import DecisionTreeClassifier # Instancia del modelo tree_car = DecisionTreeClassifier(random_state=42)

from sklearn.model_selection import GridSearchCV # parametros del decision tree param_grid = {'criterion': ['gini', 'entropy'], 'max_depth': [2, 3, 4, 5]}

# Realizar la búsqueda de hiperparámetros utilizando GridSearchCV grid_search = GridSearchCV(tree_car, param_grid=param_grid, cv=10, return_train_score=True) grid_search.fit(X_train_car, y_train_car) # Imprimir los resultados print("Mejores hiperparámetros encontrados:") print(grid_search.best_params_) print("Mejor puntuación de validación cruzada:") print(grid_search.best_score_)

# Modelo decision tree con parametros optimizados best_tree_car = grid_search.best_estimator_

# Predecimos Y y_train_pred_tree_car = best_tree_car.predict(X_train_car) y_test_pred_tree_car = best_tree_car.predict(X_test_car)

from sklearn.metrics import accuracy_score, classification_report

# Calculo el accuract en train train_acc = accuracy_score(y_true=y_test_car,y_pred=y_test_pred_tree_car) # Calculo el accuract en test test_acc = accuracy_score(y_true=y_train_car,y_pred=y_train_pred_tree_car)

print("El accuracy en train es:",train_acc) print("El accuracy en test es:",test_acc)

print(classification_report(y_test_car,y_test_pred_tree_car))

feature_scores_car = pd.DataFrame(pd.Series(best_tree_car.feature_importances_, index=X_train_car.columns).sort_values(ascending=False)).T plt.figure(figsize=(12,6)) sns.barplot(data=feature_scores_car) for index, value in enumerate(feature_scores_car.values.flatten()): plt.annotate(f'{value:.2f}', xy=(index, value), ha='center', va='bottom') plt.title("Factores clave en la predicción de la calidad de un automovil") plt.show() pd.DataFrame(feature_scores_car.T)

plt.figure(figsize = (12,8)) from sklearn import tree tree.plot_tree(best_tree_car.fit(X_train_car, y_train_car));

import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns sns.set(style='whitegrid', context='notebook')

df_diabetes = pd.read_csv('pima-indians-diabetes.csv', header=None, sep=',') df_columns = np.array(['preg','plas', 'pres', 'skin', 'test', 'mass', 'pedi', 'age', 'Class']) df_diabetes.columns = df_columns

# Visualizacion del DataFrame df_diabetes.head()

# Visualizacion del shape del DF df_diabetes.shape

# Tipos de datos df_diabetes.info()

# No hay datos duplicados en el dataset df_diabetes.drop_duplicates(inplace=True) df_diabetes

import missingno as miss miss.matrix(df_diabetes);

df_diabetes.iloc[:, 1:6].replace(to_replace=[0], value=np.nan).isna().sum().reset_index(name = 'missing_values').rename(columns={"index": "variable"}).assign( percentage = lambda df_reset: df_reset.missing_values / len(df_diabetes) * 100)

plt.figure(figsize=(7,7)) labels, counts = np.unique(df_diabetes.Class, return_counts=True) plt.pie(counts, autopct='%1.1f%%',labels=labels) plt.legend({'Diabetes Negativo','Diabetes positivo'}) plt.title('Proporcion de diabetes') plt.show() print(df_diabetes.Class.value_counts())

# Separamos en X e Y X_diabetes = df_diabetes.drop('Class',axis=1) y_diabetes = df_diabetes.Class

# Importamos las librerias para entrenamiento y testeo from sklearn.model_selection import train_test_split X_train_diabetes, X_test_diabetes, y_train_diabetes, y_test_diabetes = train_test_split(X_diabetes,y_diabetes, test_size=.20, random_state=42)

X_train_diabetes.shape,y_train_diabetes.shape

# Bosque aleatorio from sklearn.ensemble import RandomForestClassifier # Instancia del modelo rfc_diabetes = RandomForestClassifier(random_state=42)

# Optimizacion de parametros from sklearn.model_selection import GridSearchCV # Definir los hiperparámetros y sus posibles valores param_grid = { 'n_estimators': [10,25,50], 'max_depth' : [5,10,15], 'criterion' : ['mse', 'mae', 'gini', 'entropy', 'log_loss'], 'min_samples_split': [2,4,6], 'min_samples_leaf': [1,2,4], }

# Crear el objeto GridSearchCV grid_search = GridSearchCV(estimator=rfc_diabetes, param_grid=param_grid, cv=5, scoring='accuracy')

# Ajustar el modelo con GridSearchCV grid_search.fit(X_train_diabetes, y_train_diabetes) # Obtener el modelo con el mejor rendimiento best_model_diabetes = grid_search.best_estimator_

# Mejores parametros del modelo grid_search.best_params_

from sklearn.metrics import accuracy_score, classification_report y_train_pred_diabetes = best_model_diabetes.predict(X_train_diabetes) y_test_pred_diabetes = best_model_diabetes.predict(X_test_diabetes) accuracy_train_diabetes = accuracy_score(y_train_diabetes,y_train_pred_diabetes) accuracy_test_diabetes = accuracy_score(y_test_diabetes,y_test_pred_diabetes)

print(accuracy_train_diabetes) print(accuracy_test_diabetes)

print(classification_report(y_test_diabetes, y_test_pred_diabetes))

feature_scores_diabetes = pd.DataFrame(pd.Series(best_model_diabetes.feature_importances_, index=X_train_diabetes.columns).sort_values(ascending=False)).T plt.figure(figsize=(12,6)) sns.barplot(data=feature_scores_diabetes) for index, value in enumerate(feature_scores_diabetes.values.flatten()): plt.annotate(f'{value:.2f}', xy=(index, value), ha='center', va='bottom') plt.title("Factores clave en la predicción de diabetes positivo en pima indians") plt.show() pd.DataFrame(feature_scores_diabetes).T

from sklearn.model_selection import cross_val_score, KFold k = 5 cv = KFold(n_splits=k, shuffle=True, random_state=42) scores = cross_val_score(best_model_diabetes, X_train_diabetes, y_train_diabetes, cv=cv, scoring='accuracy') print("") print("Accuracy scores for each fold:", scores*100) print("Mean accuracy: ", scores.mean()*100) print("Standard deviation: ", scores.std()*100)

from sklearn.ensemble import RandomForestClassifier import matplotlib.pyplot as plt from sklearn import tree plt.figure(figsize=(30, 30)) # Obtener un árbol aleatorio del Random Forest tree_index = 0 # Índice del árbol deseado Tree = best_model_diabetes.estimators_[tree_index] # Visualizar el árbol utilizando plot_tree tree.plot_tree(Tree, feature_names=X_train_diabetes.columns, filled=True) plt.show()

X_car

y_car

X_train_car,X_test_car,y_train_car,y_test_car = train_test_split(X_car,y_car, random_state=42)

rfc_car = RandomForestClassifier(n_estimators=5,random_state=42)

rfc_car.fit(X_train_car,y_train_car)

y_pred_test_car = rfc_car.predict(X_test_car) y_pred_train_car = rfc_car.predict(X_train_car)

from sklearn.metrics import accuracy_score, classification_report, confusion_matrix

print(accuracy_score(y_train_car,y_pred_train_car)) print(accuracy_score(y_test_car,y_pred_test_car))

print(classification_report(y_test_car,y_pred_test_car))

feature_scores_car = pd.DataFrame(pd.Series(rfc_car.feature_importances_, index=X_train_car.columns).sort_values(ascending=False)).T plt.figure(figsize=(12,6)) sns.barplot(data=feature_scores_car) for index, value in enumerate(feature_scores_car.values.flatten()): plt.annotate(f'{value:.2f}', xy=(index, value), ha='center', va='bottom') plt.title("Factores clave en la predicción de la calidad de un automovil") plt.show() pd.DataFrame(feature_scores_car).T