import pandas as pd
import numpy as np
import tensorflow as tf
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
import joblib

# 1. Data Acquisition (Using the standard insurance dataset)
url = "https://raw.githubusercontent.com/stedy/Machine-Learning-with-R-datasets/master/insurance.csv"
df = pd.read_csv(url)

# 2. Preparation: Encoding Categorical Data
df['sex'] = df['sex'].map({'female': 0, 'male': 1})
df['smoker'] = df['smoker'].map({'no': 0, 'yes': 1})
# One-hot encode the 'region' column
df = pd.get_dummies(df, columns=['region'])

# 3. Splitting Features and Target
X = df.drop('charges', axis=1)
y = df['charges']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

# 4. Feature Scaling
scaler = StandardScaler()
X_train_scaled = scaler.fit_transform(X_train)
X_test_scaled = scaler.transform(X_test)

# Save the scaler for use in the app
joblib.dump(scaler, 'scaler.pkl')

# 5. Model Building (ANN Architecture)
model = tf.keras.Sequential([
    tf.keras.layers.Dense(128, activation='relu', input_shape=(X_train_scaled.shape[1],)),
    tf.keras.layers.Dropout(0.2),
    tf.keras.layers.Dense(64, activation='relu'),
    tf.keras.layers.Dense(32, activation='relu'),
    tf.keras.layers.Dense(1) # Regression output
])

# 6. Compilation & Training
model.compile(optimizer='adam', loss='mae', metrics=['mse'])
model.fit(X_train_scaled, y_train, epochs=150, batch_size=32, validation_split=0.1, verbose=1)

# 7. Save the Model
model.save('insurance_model.h5')
print("Model and Scaler saved successfully!")