Goal:
- Explore the data (handle missing values etc).
- Analysis per store type and correlational analysis of stores activity.
- Perform extensive Time Series Analysis (seasonal decomposition, trends, autocorrelation).
- Predict next 6 weeks of sales using Prophet (Facebook methodology).
This notebook mainly focuses on the Time Series Analysis. An important topic yet not covered. I use then new methodology Prophet, recently introduced by Facebook, to predict next 6 week of sales. This methodology has a cool feature of modeling for holidays. Finally, right at the end, I also discussadvantages and drawbacks of forecasting with Seasonal ARIMA and Prophet.
As it usually goes, we start with the Exploratory Data Analysis of the main metrics revealing present trends and patterns in the data, giving a solid foundation for the further (causal) analysis.