import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import warnings
warnings.filterwarnings('ignore')
np.random.seed(42)

n_transactions = 1000
n_customers = 200
df = pd.DataFrame({
    'transaction_id': range(1, n_transactions + 1),
    'customer_id': np.random.choice([f'CUST_{i:04d}' for i in range(1, n_customers + 1)], n_transactions),
    'amount': np.round(np.random.normal(150, 60, n_transactions), 2),
    'transaction_type': np.random.choice(['Debit', 'Credit'], n_transactions),
    'channel': np.random.choice(['ATM', 'Online', 'Branch', 'POS'], n_transactions),
    'date': pd.date_range(start='2024-01-01', periods=n_transactions, freq='h')
})
print(df.shape)
print(df.head(3))

# Basic aggregation: sum transaction amounts per day
df['date_day'] = df['date'].dt.date
daily_amounts = df.groupby('date_day')['amount'].sum().reset_index()
print(daily_amounts.head())

# Beginner Example 1: Simple 3-day moving average
daily_amounts['MA_3'] = daily_amounts['amount'].rolling(window=3).mean()
print(daily_amounts[['amount','MA_3']].head(5))

# Beginner Example 2: 7-day moving average for weekly smoothing
daily_amounts['MA_7'] = daily_amounts['amount'].rolling(window=7).mean()
print(daily_amounts[['amount','MA_7']].head(10))

# Beginner Example 3: Visualizing daily amounts vs 7-day moving average
plt.figure(figsize=(10,5))
plt.plot(daily_amounts['date_day'], daily_amounts['amount'], label='Daily Total', alpha=0.7)
plt.plot(daily_amounts['date_day'], daily_amounts['MA_7'], label='7-day MA', color='red', linewidth=2)
plt.xlabel('Date')
plt.ylabel('Amount')
plt.title('Daily Transaction Totals and 7-day Moving Average')
plt.legend()
plt.tight_layout()
plt.show()

# Intermediate Example 1: 14-day moving average for trend analysis
daily_amounts['MA_14'] = daily_amounts['amount'].rolling(window=14).mean()
print(daily_amounts[['amount','MA_14']].tail(10))

# Intermediate Example 2: Centered moving average
daily_amounts['MA_7_centered'] = daily_amounts['amount'].rolling(window=7, center=True).mean()
print(daily_amounts[['amount','MA_7','MA_7_centered']].head(10))

# Intermediate Example 3: Exponential Weighted Moving Average (EWMA)
daily_amounts['EWMA_7'] = daily_amounts['amount'].ewm(span=7, adjust=False).mean()
print(daily_amounts[['amount','EWMA_7']].head(10))

# Intermediate Example 4: Plot all types of smoothing
plt.figure(figsize=(12,6))
plt.plot(daily_amounts['date_day'], daily_amounts['amount'], label='Raw', alpha=0.5)
plt.plot(daily_amounts['date_day'], daily_amounts['MA_7'], label='7d MA', color='red')
plt.plot(daily_amounts['date_day'], daily_amounts['MA_14'], label='14d MA', color='green', linestyle='--')
plt.plot(daily_amounts['date_day'], daily_amounts['EWMA_7'], label='7d EWMA', color='orange')
plt.xlabel('Date')
plt.ylabel('Amount')
plt.title('Comparison of Moving Averages - Banking Transactions')
plt.legend()
plt.tight_layout()
plt.show()

# Advanced Example 1: Handling outliers in banking data
daily_amounts['amount_no_outlier'] = np.where(
    (daily_amounts['amount'] > daily_amounts['amount'].quantile(0.99)) |
    (daily_amounts['amount'] < daily_amounts['amount'].quantile(0.01)),
    np.nan, daily_amounts['amount']
)
daily_amounts['MA_7_no_outlier'] = daily_amounts['amount_no_outlier'].rolling(window=7).mean()
print(daily_amounts[['amount','MA_7','MA_7_no_outlier']].tail(10))

# Advanced Example 2: Using groupby for customer-level smoothing
df['date_day'] = df['date'].dt.date
cust_daily = df.groupby(['customer_id','date_day'])['amount'].sum().reset_index()
cust_daily['MA_3_by_cust'] = cust_daily.groupby('customer_id')['amount'].rolling(window=3).mean().reset_index(0,drop=True)
print(cust_daily[cust_daily['customer_id']=='CUST_0001'].head(6))

# Advanced Example 3: Detecting anomalies: flagging points far from the moving average
cust_daily['deviation'] = cust_daily['amount'] - cust_daily['MA_3_by_cust']
cust_daily['anomaly_flag'] = np.abs(cust_daily['deviation']) > 2 * cust_daily.groupby('customer_id')['amount'].transform('std')
anomalies = cust_daily[cust_daily['anomaly_flag']]
print(anomalies.head(5))

# Handling NaNs from insufficient data
try:
    test_df = pd.DataFrame({'val':[1,2,np.nan,4,5]})
    test_df['MA_3'] = test_df['val'].rolling(window=3).mean()
    print(test_df)
except Exception as e:
    print('Error:', e)

# Defensive smoothing: fill NaNs and proceed
safe_df = test_df.copy()
safe_df['val_filled'] = safe_df['val'].fillna(method='ffill')
safe_df['MA_3_filled'] = safe_df['val_filled'].rolling(window=3).mean()
print(safe_df)

# Common pattern: wrapper function for moving average
def compute_moving_average(series, window, center=False):
    return series.rolling(window=window, center=center).mean()
 
daily_amounts['MA_func'] = compute_moving_average(daily_amounts['amount'], 5)
print(daily_amounts[['amount','MA_func']].head(10))

# End-to-end Example: Monthly cash flow smoothing for operational planning
df['month'] = df['date'].dt.to_period('M')
monthly_amounts = df.groupby('month')['amount'].sum().reset_index()
monthly_amounts['MA_3_month'] = monthly_amounts['amount'].rolling(window=3).mean()
monthly_amounts['EWMA_3_month'] = monthly_amounts['amount'].ewm(span=3, adjust=False).mean()
print(monthly_amounts)

# Visualizing monthly smoothing results
plt.figure(figsize=(8,4))
plt.plot(monthly_amounts['month'].astype(str), monthly_amounts['amount'], label='Monthly Total', marker='o')
plt.plot(monthly_amounts['month'].astype(str), monthly_amounts['MA_3_month'], label='3-MA', marker='s')
plt.plot(monthly_amounts['month'].astype(str), monthly_amounts['EWMA_3_month'], label='3-EWMA', marker='^')
plt.xlabel('Month')
plt.ylabel('Amount')
plt.title('Monthly Transaction Totals and Smoothing')
plt.legend()
plt.tight_layout()
plt.show()