Predict Demand Spikes with statsforecast and scikit-learn

                      
                     
import os
import pandas as pd
from statsforecast import StatsForecast
from statsforecast.models import AutoARIMA
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error

# Configuration
DATA_PATH = os.getenv('DATA_PATH', 'data/demand_data.csv')
FORECAST_HORIZON = 7  # Days to forecast

# Load data
try:
    data = pd.read_csv(DATA_PATH)
    data['date'] = pd.to_datetime(data['date'])
    data.set_index('date', inplace=True)
except Exception as e:
    raise RuntimeError(f'Failed to load data: {str(e)}')

# Prepare training and test sets
train, test = train_test_split(data, test_size=0.2, shuffle=False)

# Initialize StatsForecast with AutoARIMA model
model = StatsForecast(models=[AutoARIMA()], freq='D')

# Fit model and predict
try:
    model.fit(train)
    forecast = model.predict(steps=FORECAST_HORIZON)
except Exception as e:
    raise RuntimeError(f'Model training or prediction failed: {str(e)}')

# Evaluate predictions
try:
    mse = mean_squared_error(test['demand'][:FORECAST_HORIZON], forecast)
    print(f'Mean Squared Error: {mse}')
except Exception as e:
    raise RuntimeError(f'Error in evaluation: {str(e)}')

if __name__ == '__main__':
    print('Demand prediction completed successfully.')