Time Series Forecasting Part 3 - Vector Auto Regression
Go to main | Course Page
Section 1: VAR
- What is VAR and When can we use it?
- Download Resources
- Intuition behind VAR model formula
- Building VAR model in Python
- Import Dataset
- Visualize the time series
- Testing Causation using Granger’s Causality Test
- Cointegration Test
- Split the series into training and testing data
- Check for the stationary and make Time Series Stationary
- How to select the order (p) of VAR model
- Train the VAR model of selected order
- Check for serial correlation of residual (errors) using Durbin Watson
Statistics
- How to forecast VAR model using stats model
- Invert the transformation to get the real forecast
- Plot of Forecast vs Actual
- Evaluate the forecast
- VARMA Model
- VARMAX with Exogenous Data
- Auto ARIMA on VARMA for Model Selection
Section 2: Practical Advise for succeeding in Time Series Project
- Align your forecast with working group ofshareholders
- Directions matters more
- Choice of Evaluation Matric
- Set realistic success criteria
- Educate your Stakeholders (Though they might not always listen)
- Explain your forecast
- Explain drop in the forecast after going live. Have root cause analysis
mechanism involving domain expert