# Always start by suppressing warnings
import warnings
import numpy as np
np.random.seed(42)
warnings.filterwarnings("ignore")

# Data setup: Load a real world time series dataset (ambient temperature with system failure events)
import pandas as pd
url = 'https://raw.githubusercontent.com/numenta/NAB/master/data/realKnownCause/ambient_temperature_system_failure.csv'
df = pd.read_csv(url)
print(df.shape)
print(df.head(3))

# Convert time column to datetime and sort data
df['timestamp'] = pd.to_datetime(df['timestamp'])
df = df.sort_values('timestamp').reset_index(drop=True)
print(df.dtypes[:2])
print(df.head(3))

# Plot the temperature data over time
import matplotlib.pyplot as plt
plt.figure(figsize=(12,4))
plt.plot(df['timestamp'], df['value'], color='skyblue')
plt.title('Ambient Temperature Over Time')
plt.xlabel('Time')
plt.ylabel('Temperature')
plt.show()

# Describe the temperature data to find basic statistics
stats = df['value'].describe()
print(stats)

# Check for missing temperature values
missing = df['value'].isna().sum()
print(f"Missing temperature readings: {missing}")

# Remove any missing values just in case
df = df.dropna(subset=['value']).reset_index(drop=True)
print(df.shape)

# Mark anomalies using the mean and standard deviation rule
mean = df['value'].mean()
std = df['value'].std()
threshold = 3 * std
df['anomaly'] = abs(df['value'] - mean) > threshold
print(df['anomaly'].value_counts())

# Plot anomalies on the time series chart
plt.figure(figsize=(12,4))
plt.plot(df['timestamp'], df['value'], label='Temperature', color='lightgray')
plt.scatter(df.loc[df['anomaly'],'timestamp'], df.loc[df['anomaly'],'value'],
            color='red', label='Anomaly', zorder=5)
plt.legend()
plt.title('Temperature with Anomalies Highlighted')
plt.xlabel('Time')
plt.ylabel('Temperature')
plt.show()

# Detect anomalies using moving average and moving standard deviation
window = 50  # adjust as needed
df['rolling_mean'] = df['value'].rolling(window).mean()
df['rolling_std'] = df['value'].rolling(window).std()
df['anomaly_ma'] = abs(df['value'] - df['rolling_mean']) > 3 * df['rolling_std']
print(df['anomaly_ma'].value_counts())

# Visualize anomalies found by the moving average method
plt.figure(figsize=(12,4))
plt.plot(df['timestamp'], df['value'], color='lightgray', label='Temperature')
plt.plot(df['timestamp'], df['rolling_mean'], color='blue', alpha=0.7, label='Moving Average')
plt.scatter(df.loc[df['anomaly_ma'], 'timestamp'], df.loc[df['anomaly_ma'], 'value'],
            color='orange', label='Anomaly (MA)', zorder=5)
plt.legend()
plt.title('Moving Average Anomaly Detection')
plt.xlabel('Time')
plt.ylabel('Temperature')
plt.show()

# Mini project: Try changing the rolling window size
window_size = int(input("Choose a new window size (like 25, 50, 100): "))
df['roll_mean2'] = df['value'].rolling(window_size).mean()
df['roll_std2'] = df['value'].rolling(window_size).std()
df['anomaly_custom'] = abs(df['value'] - df['roll_mean2']) > 3 * df['roll_std2']
plt.figure(figsize=(12,4))
plt.plot(df['timestamp'], df['value'], color='gray')
plt.plot(df['timestamp'], df['roll_mean2'], color='green', alpha=0.7)
plt.scatter(df.loc[df['anomaly_custom'],'timestamp'], df.loc[df['anomaly_custom'],'value'],
            c='magenta', label='Custom Anomaly', zorder=5)
plt.legend()
plt.title(f'Moving Average Window: {window_size}')
plt.show()

# Save detected anomalies to a CSV file for review
anomalies = df.loc[df['anomaly_custom'], ['timestamp', 'value']]
anomalies.to_csv('anomalies_detected.csv', index=False)
print(f"Saved {anomalies.shape[0]} anomalies to anomalies_detected.csv")