pip install darts

import pandas as pd
import numpy as np
from darts import TimeSeries
from darts.models import ExponentialSmoothing# Create sample data
date_rng = pd.date_range(start='2023-01-01', periods=100, freq='D')
data = np.random.randn(100) * 10 + 50  # Random values around 50

# Create a DataFrame
df = pd.DataFrame({'Date': date_rng, 'Value': data})

# Convert to TimeSeries
train = TimeSeries.from_dataframe(df, 'Date', 'Value')

import pandas as pd
from darts import TimeSeries
from darts.models import ExponentialSmoothing
from darts.metrics import mae, rmse
import matplotlib.pyplot as plt

# Grab the sales log
df = pd.read_csv("AirPassengers.csv", delimiter=",")

# Turn it into something Darts can chew on
series = TimeSeries.from_dataframe(df, "Month", "#Passengers")

# Split it: most for baking (training), the last 36 months to taste-test (validation)
train = series[:-36]  # First 108 months (Jan 1949 - Dec 1957)
val = series[-36:]   # Last 36 months (Jan 1958 - Dec 1960)

# Create and fit an Exponential Smoothing model
model = ExponentialSmoothing(seasonal_periods=12)  # 12 months for yearly seasonality
model.fit(train)

# Predict the next 36 months
prediction = model.predict(n=36)

# Evaluate the predictions
mae_error = mae(val, prediction)
rmse_error = rmse(val, prediction)
print(f"Mean Absolute Error (MAE): {mae_error:.2f}")
print(f"Root Mean Squared Error (RMSE): {rmse_error:.2f}")

# Visualize the results
plt.figure(figsize=(12, 6))
train.plot(label="Training Data")
val.plot(label="Actual Validation Data")
prediction.plot(label="Predicted Validation Data", linestyle="--")
plt.title("AirPassengers Forecasting")
plt.xlabel("Month")
plt.ylabel("#Passengers")
plt.legend()
plt.grid(True)
plt.show()

# Optional: Print a few values for inspection
print("\nFirst 5 training values:")
print(train[:5].values())
print("\nFirst 5 validation values:")
print(val[:5].values())
print("\nFirst 5 predicted values:")
print(prediction[:5].values())

# Create sample data
date_rng = pd.date_range(start='2023-01-01', periods=100, freq='D')
data = np.random.randn(100) * 10 + 50  # Random values around 50

# Create a DataFrame
df = pd.DataFrame({'Date': date_rng, 'Value': data})

# Convert to TimeSeries
train = TimeSeries.from_dataframe(df, 'Date', 'Value')

# Initialize the model
model = ExponentialSmoothing()

# Fit the model
model.fit(train)

# Predict the next 10 time steps
forecast = model.predict(10)

# Print the forecast
print(forecast)

import matplotlib.pyplot as plt

# Plot the original data
train.plot(label="Training Data", color='blue')

# Plot the forecast
forecast.plot(label="Forecast", color='red')

# Customize the plot
plt.title("Exponential Smoothing Forecast")
plt.xlabel("Date")
plt.ylabel("Value")
plt.legend()
plt.show()

# Bake up 36 months’ worth, with some wiggle room
prediction = model.predict(n=len(val), num_samples=1000)

import numpy as np

# Assuming `prediction` is the output of model.predict()
mean_forecast = np.mean(prediction, axis=1)  # Mean prediction per time step
lower_bound = np.percentile(prediction, 5, axis=1)  # 5th percentile (lower uncertainty bound)
upper_bound = np.percentile(prediction, 95, axis=1)  # 95th percentile (upper uncertainty bound)

import matplotlib.pyplot as plt

time_steps = range(len(val))  # X-axis (assumed same length as validation set)

plt.figure(figsize=(12, 6))

# Plot the mean forecast
plt.plot(time_steps, mean_forecast, label="Mean Prediction", color='blue')

# Plot the confidence interval (shaded region)
plt.fill_between(time_steps, lower_bound, upper_bound, color='blue', alpha=0.2, label="90% Confidence Interval")

# Plot the actual validation data for comparison (if available)
plt.plot(time_steps, val, label="Actual Data", color='black', linestyle='dashed')

plt.xlabel("Time")
plt.ylabel("Predicted Value")
plt.legend()
plt.title("Probabilistic Forecast with Confidence Intervals")
plt.show()