This intermediate training course focuses on building robust forecasting models using ARIMA, SARIMA, and seasonal decomposition techniques. Designed for professionals working with periodic and structured time series data, the professional course covers parameter tuning, trend-seasonality separation, and multi-step forecasting. The program emphasizes hands-on modeling using Python’s Statsmodels and pmdarima libraries.
Case-driven forecasting labs using real datasets
Step-by-step model tuning and residual diagnostics
Visualization of seasonality and forecasting intervals
Final forecasting challenge and discussion
Recognize and analyze seasonal patterns in time series
Apply seasonal decomposition (additive and multiplicative models)
Build ARIMA models and tune parameters using AIC/BIC
Extend ARIMA to SARIMA for seasonally adjusted forecasting
Perform model diagnostics and residual analysis
Forecast and backtest time series with multiple steps
Select the best model using statistical evaluation
Day 1
Session 1: Seasonality and Trend in Time Series
Decomposition of time series
Additive vs multiplicative models
Visual analysis of seasonal cycles
Session 2: Classical Decomposition and STL
Moving averages and smoothing
Seasonal-Trend decomposition (STL)
Plotting components with Statsmodels
Day 2
Session 3: ARIMA Recap and Model Tuning
Refresher on ARIMA (p, d, q)
Manual vs auto ARIMA (using pmdarima)
Using AIC/BIC for model selection
Session 4: SARIMA – Modeling Seasonality with ARIMA
SARIMA structure (p, d, q) × (P, D, Q, s)
Model fitting and interpretation
Seasonal lags and period selection
Day 3
Session 5: Residual Diagnostics and Forecast Validation
Ljung-Box test and autocorrelation of residuals
Forecast accuracy metrics: MAPE, RMSE
Confidence intervals and prediction bounds
Session 6: Advanced Forecasting Techniques
Multi-step forecasting
Rolling forecasts and sliding window validation
Dealing with anomalies and missing values
Day 4
Session 7: Applied Forecasting Lab
Forecasting monthly/quarterly datasets
SARIMA tuning and evaluation
Backtesting predictions
Session 8: Final Review and Transition to Advanced Topics
When ARIMA/SARIMA is not enough
Overview of Prophet and LSTM-based forecasting
Best practices for long-term forecasting in business
We are open to customizing this program to align with your specific learning objectives. If your team has particular goals or areas they wish to focus on, we would be happy to tailor the course outline to meet those needs and ensure the program supports the achievement of your desired outcomes.