import warnings
warnings.filterwarnings("ignore")
import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

# Data setup
titanic = sns.load_dataset("titanic")
print("Shape:", titanic.shape)
titanic.head()

# Let''s check missing values in the main columns we will use
print(titanic[["survived", "sex", "age", "pclass"]].isnull().sum())

# For simplicity, let''s drop rows with missing age
titanic_clean = titanic.dropna(subset=["age"]).copy()
print("Rows after dropping missing ages:", len(titanic_clean))

# Convert categorical variables to numeric codes
titanic_clean["sex_code"] = titanic_clean["sex"].map({"male": 0, "female": 1})
titanic_clean["pclass_code"] = titanic_clean["pclass"]

# Let''s see how probability and odds work
p = 0.8
odds = p / (1 - p)
log_odds = np.log(odds)
print(f"If probability is {p}, odds are {odds:.2f}, and log-odds are {log_odds:.2f}")

from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import train_test_split

# Set up X and y for modeling
X = titanic_clean[["sex_code", "pclass_code", "age"]]
y = titanic_clean["survived"]

# Split into train and test (80/20)
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# Fit the logistic regression model
lr = LogisticRegression(solver="lbfgs")
lr.fit(X_train, y_train)

# Look at model coefficients
features = ["sex_code", "pclass_code", "age"]
for name, coef in zip(features, lr.coef_[0]):
    print(f"{name:10s}: {coef:+.3f}")
    

# Calculate odds ratios by exponentiating coefficients
odds_ratios = np.exp(lr.coef_[0])
for name, oratio in zip(features, odds_ratios):
    print(f"{name:10s}: {oratio:.2f}")
    

# Predict survival probabilities for a sample passenger
example = pd.DataFrame({"sex_code": [0, 1], "pclass_code": [3, 1], "age": [25, 25]})
probs = lr.predict_proba(example)[:, 1]
print(f"Male, 3rd class, age 25 survives: {probs[0]:.2f}")
print(f"Female, 1st class, age 25 survives: {probs[1]:.2f}")

# See how odds ratio changes with age
age_diff = 10
ex1 = pd.DataFrame({"sex_code": [0], "pclass_code": [2], "age": [20]})
ex2 = pd.DataFrame({"sex_code": [0], "pclass_code": [2], "age": [20 + age_diff]})
prob1 = lr.predict_proba(ex1)[:, 1][0]
prob2 = lr.predict_proba(ex2)[:, 1][0]
change = prob2 / prob1 if prob1 > 0 else np.nan
print(f"A 10-year increase in age changes survival chance from {prob1:.2f} to {prob2:.2f}. Ratio: {change:.2f}")

# What if we use all test data? Check predicted vs. actual
y_pred = lr.predict(X_test)
accuracy = np.mean(y_pred == y_test)
print(f"Model accuracy on held-out data: {accuracy:.2%}")

# Practice: Try your own inputs!
sex = input("Enter 0 for male, 1 for female: ")
pclass = input("Enter 1 for 1st, 2 for 2nd, or 3 for 3rd class: ")
age = input("Enter passenger age: ")
your_data = pd.DataFrame({
    "sex_code": [int(sex)],
    "pclass_code": [int(pclass)],
    "age": [float(age)]
})
prob = lr.predict_proba(your_data)[:, 1][0]
print(f"Your survival chance: {100*prob:.1f}%")