import os import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns import tensorflow as tf from tensorflow import keras from sklearn.model_selection import train_test_split from sklearn.preprocessing import StandardScaler, OneHotEncoder from sklearn.impute import SimpleImputer from sklearn.compose import ColumnTransformer from sklearn.pipeline import Pipeline from sklearn.metrics import (confusion_matrix, classification_report, roc_auc_score, roc_curve, accuracy_score)

df = pd.read_csv('train.csv') df.head()

df2 = df.copy() cols_drop = ['Ticket','Name','Cabin'] for c in cols_drop: if c in df2.columns: df2 = df2.drop(columns=c) df2['FamilySize'] = df2['SibSp'] + df2['Parch'] + 1 df2['IsAlone'] = (df2['FamilySize'] == 1).astype(int) df2['Pclass'] = df2['Pclass'].astype(str) df2.head()

target = 'Survived' X = df2.drop(columns=[target]) y = df2[target].values # Columnas num_features = ['Age','Fare','SibSp','Parch','FamilySize'] cat_features = ['Sex','Embarked','Pclass'] # Pipelines num_pipe = Pipeline([ ('imputer', SimpleImputer(strategy='median')), ('scaler', StandardScaler()) ]) cat_pipe = Pipeline([ ('imputer', SimpleImputer(strategy='most_frequent')), ('onehot', OneHotEncoder(handle_unknown='ignore', sparse=False)) ]) preprocessor = ColumnTransformer([ ('num', num_pipe, num_features), ('cat', cat_pipe, cat_features) ])

SEED = 42 # Split the data X_train, X_val, y_train, y_val = train_test_split( X, y, test_size=0.2, random_state=SEED, stratify=y) X_train_proc = preprocessor.fit_transform(X_train) X_val_proc = preprocessor.transform(X_val) X_train_proc = X_train_proc.astype(np.float32) X_val_proc = X_val_proc.astype(np.float32) X_train_proc.shape, X_val_proc.shape

def build_model(input_dim, n_hidden_layers=2, n_neurons=16, activation='relu', learning_rate=0.001, dropout_rate=0.0): inputs = keras.Input(shape=(input_dim,)) x = inputs for i in range(n_hidden_layers): x = keras.layers.Dense(n_neurons, activation=activation)(x) if dropout_rate and dropout_rate > 0: x = keras.layers.Dropout(dropout_rate)(x) outputs = keras.layers.Dense(1, activation='sigmoid')(x) model = keras.Model(inputs, outputs) optimizer = keras.optimizers.Adam(learning_rate=learning_rate) model.compile(optimizer=optimizer, loss='binary_crossentropy', metrics=['accuracy', tf.keras.metrics.AUC(name='auc')]) return model

# Define the model before using it input_dim = X_train_proc.shape[1] # Number of features in the input model = build_model(input_dim=input_dim, n_hidden_layers=2, n_neurons=16, activation='relu', learning_rate=0.001, dropout_rate=0.2) # Train the model epochs = 60 batch_size = 32 callbacks = [ keras.callbacks.EarlyStopping(monitor='val_auc', mode='max', patience=6, restore_best_weights=True) ] history = model.fit( X_train_proc, y_train, validation_data=(X_val_proc, y_val), epochs=epochs, batch_size=batch_size, callbacks=callbacks, verbose=2 )

plt.figure(figsize=(8,5)) plt.plot(history.history['loss'], label='train_loss') plt.plot(history.history['val_loss'], label='val_loss') plt.xlabel('Época') plt.ylabel('Loss (binary_crossentropy)') plt.title('Loss por época') plt.legend() plt.grid(True) plt.show()

y_val_proba = model.predict(X_val_proc).ravel() y_val_pred = (y_val_proba >= 0.5).astype(int) print("Accuracy:", accuracy_score(y_val, y_val_pred)) print("ROC AUC:", roc_auc_score(y_val, y_val_proba)) print("\nClassification report:\n", classification_report(y_val, y_val_pred)) cm = confusion_matrix(y_val, y_val_pred) sns.heatmap(cm, annot=True, fmt='d', cmap='Blues') plt.xlabel('Predicho') plt.ylabel('Real') plt.title('Matriz de confusión') plt.show() fpr, tpr, _ = roc_curve(y_val, y_val_proba) plt.figure(figsize=(6,5)) plt.plot(fpr, tpr, label=f'AUC = {roc_auc_score(y_val, y_val_proba):.3f}') plt.plot([0,1],[0,1],'k--') plt.xlabel('FPR') plt.ylabel('TPR') plt.title('ROC Curve') plt.legend() plt.grid(True) plt.show()

from itertools import product results = [] neuron_list = [16, 32] layers_list = [1, 2] activation_list = ['relu','tanh'] lr_list = [1e-3, 1e-4] batch_list = [32] for neurons, layers, act, lr, batch in product(neuron_list, layers_list, activation_list, lr_list, batch_list): print(f"Entrenando: neurons={neurons}, layers={layers}, act={act}, lr={lr}, batch={batch}") m = build_model(input_dim, n_hidden_layers=layers, n_neurons=neurons, activation=act, learning_rate=lr) h = m.fit(X_train_proc, y_train, validation_data=(X_val_proc, y_val), epochs=30, batch_size=batch, callbacks=[keras.callbacks.EarlyStopping(monitor='val_auc', mode='max', patience=5, restore_best_weights=True)], verbose=0) best_val_auc = max(h.history['val_auc']) best_val_loss = min(h.history['val_loss']) results.append({ 'neurons': neurons, 'layers': layers, 'activation': act, 'lr': lr, 'batch': batch, 'val_auc': best_val_auc, 'val_loss': best_val_loss }) res_df = pd.DataFrame(results).sort_values(by='val_auc', ascending=False) res_df.head(10)

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, confusion_matrix, classification_report, roc_auc_score, roc_curve import seaborn as sns import matplotlib.pyplot as plt # Predicciones y_val_proba = model.predict(X_val_proc).ravel() y_val_pred = (y_val_proba >= 0.5).astype(int) # Métricas numéricas print("Accuracy:", accuracy_score(y_val, y_val_pred)) print("Precision:", precision_score(y_val, y_val_pred)) print("Recall:", recall_score(y_val, y_val_pred)) print("F1-score:", f1_score(y_val, y_val_pred)) print("ROC AUC:", roc_auc_score(y_val, y_val_proba)) print("\nClassification report:\n", classification_report(y_val, y_val_pred)) # Matriz de confusión cm = confusion_matrix(y_val, y_val_pred) sns.heatmap(cm, annot=True, fmt="d", cmap="Blues") plt.xlabel("Predicho") plt.ylabel("Real") plt.title("Matriz de confusión") plt.show() # Curva ROC fpr, tpr, _ = roc_curve(y_val, y_val_proba) plt.figure(figsize=(6,5)) plt.plot(fpr, tpr, label=f"AUC = {roc_auc_score(y_val, y_val_proba):.3f}") plt.plot([0,1],[0,1],'k--') plt.xlabel("FPR") plt.ylabel("TPR") plt.title("ROC Curve") plt.legend() plt.grid(True) plt.show()

df = pd.read_csv('test.csv') df.head()

test = pd.read_csv("test.csv") test_id = test["PassengerId"]

test2 = test.copy() test2 = test2.drop(columns=["Name","Ticket","Cabin"], errors="ignore") test2["FamilySize"] = test2["SibSp"] + test2["Parch"] + 1 test2["IsAlone"] = (test2["FamilySize"] == 1).astype(int) test2["Pclass"] = test2["Pclass"].astype(str) # 3. Aplicar el preprocesador entrenado en train test_proc = preprocessor.transform(test2) # 4. Predicciones test_proba = model.predict(test_proc).ravel() test_pred = (test_proba >= 0.5).astype(int) # 5. Crear submission submission = pd.DataFrame({ "PassengerId": test_id, "Survived": test_pred }) submission.to_csv("submission.csv", index=False) print("Archivo submission.csv creado") submission.head()