import pandas as pd
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import MinMaxScaler
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score

# ==========================================
# 1. INPUT DATA (Data Collection)
# ==========================================
# Simulate 10 customers' data (in reality, data comes from SQL Database or CSV)
# Assume 1 = Churn, 0 = Stay
data = {
    'Customer_ID': ['C001', 'C002', 'C003', 'C004', 'C005', 'C006', 'C007', 'C008', 'C009', 'C010'],
    'Name': ['Alice', 'Bob', 'Charlie', 'David', 'Eve', 'Frank', 'Grace', 'Hannah', 'Ian', 'Jack'],
    'Age': [25, 45, 35, 50, 23, 60, 32, 28, 40, 55],
    'Salary': [18000, 85000, 45000, 120000, 15000, 90000, 50000, 22000, 60000, 110000],
    'Days_Since_Last_Buy': [2, 45, 10, 5, 60, 3, 15, 40, 7, 2],
    'Total_Spend': [500, 12000, 3000, 25000, 200, 15000, 4000, 800, 5500, 30000],
    'Churn_Label': [0, 1, 0, 0, 1, 0, 0, 1, 0, 0] 
}

df = pd.DataFrame(data)

print("--- Raw Data ---

")
print(df.head(3)) # View only the first 3 rows
print("\n")

# ==========================================
# 2. DATA PREPROCESSING
# ==========================================

# --- 2.1 Feature Selection (Select only features that matter) ---


# HINT: ID and Name are Unique Identifiers and have no effect on purchasing behavior. Discard them.
# This is the simplest dimension reduction (Manual Feature Selection)
features = df.drop(['Customer_ID', 'Name', 'Churn_Label'], axis=1)
labels = df['Churn_Label']

# --- 2.2 Data Subsetting (Train/Test Split) ---


# HINT: Split data for "teaching" (Train) and "testing" (Test) to simulate real-world usage
# Randomly assign 80% to train, 20% to test
X_train, X_test, y_train, y_test = train_test_split(features, labels, test_size=0.2, random_state=42)

# --- 2.3 Normalization (Scaling data) ---


# HINT: Salary in tens of thousands vs Age in tens. If not scaled, the model will bias towards Salary
# We use Min-Max Scaler to scale everything to be between 0 and 1
scaler = MinMaxScaler()

# fit_transform on Train data (remembers Min/Max of Train)
X_train_scaled = scaler.fit_transform(X_train) 
# transform Test data (uses the same Min/Max from Train for fairness)
X_test_scaled = scaler.transform(X_test)

print("--- Data after Preprocessing (Normalized) ---

")
print(pd.DataFrame(X_train_scaled, columns=features.columns).head(3))
print("\n")

# ==========================================
# 3. DATA MINING
# ==========================================

# Choose Algorithm: Decision Tree
# Because it's easy to understand and suitable for beginners
clf = DecisionTreeClassifier(random_state=42)

# Make the model "learn" from prepared data (Machine Learning)
clf.fit(X_train_scaled, y_train)

# Try predicting the test set
y_pred = clf.predict(X_test_scaled)

# ==========================================
# 4. POSTPROCESSING & INFORMATION
# ==========================================

# --- 4.1 Evaluation ---


accuracy = accuracy_score(y_test, y_pred)
print(f"--- Output (Information) ---

")
print(f"Model Accuracy: {accuracy * 100:.2f}%")

# --- 4.2 Pattern Interpretation (See what the model knows) ---


# HINT: Look at which Feature is most important in deciding whether the customer will churn or not
feature_importance = pd.DataFrame({
    'Feature': features.columns,
    'Importance': clf.feature_importances_
}).sort_values(by='Importance', ascending=False)

print("\nFactors affecting membership churn (sorted by importance):")
print(feature_importance)

# ==========================================
# 5. VISUALIZATION (Bonus: Look at the model)
# ==========================================
import matplotlib.pyplot as plt
from sklearn.tree import plot_tree

# Set the image size (width 15 inches, height 10 inches) for clear viewing
plt.figure(figsize=(15, 10))

# Tree drawing command
plot_tree(clf, 
          feature_names=features.columns,  # Add column names for easy reading (Age, Salary)
          class_names=['Stay (0)', 'Churn (1)'], # Name the results (0=Stay, 1=Churn)
          filled=True,     # Color boxes by result group (Blue=Stay, Orange=Churn)
          rounded=True,    # Make rounded edges for beauty
          fontsize=12)     # Font size

plt.title("Decision Tree Logic for Customer Churn Prediction")
plt.show() # Display the image