MBA 8040

Assignment 2: Time Series Forecasting

Collect time series data – either quarterly or monthly or daily – with 4 or 5 seasonal cycles. That is, four or five years of data if quarterly or monthly aggregates are recorded, or 4 to 5 weeks of data if daily observations were recorded. Alternately, use the Walmart data posted on the website, and select only the last 16 quarters, starting Jan 2013.

Note that even though the data shows seasonality, and in reality all you would do is the decomposition method, I want you to demonstrate your understanding of each method learned in class. Therefore, create Time Series Forecasting models using the following three techniques:

1. Moving Average (you can choose 3, or 4 or 5 period moving average – pick any one and directly forecast using that, as though the data has no trend or seasonality)

2. Trendline (this is where you simply do a trendline with the original y and forecast using just that, as though there was no seasonality in data)

3. Decomposition (this technique assumes both trend and seasonality – in the process of doing this, you will compute moving averages, centered averages, seasonal indices, deseasonalized y, and then do a trendline on that, followed by re-seasonalizing).

Evaluate each method by computing Bias, MAD, MAPE, and MSE (or Standard Error). Which method worked best on your data? What would be your forecast for the next time period based on that method?

Deliverables:

1. Write a report in Word that summarizes your analyses. Copy any tables/charts/graphs from Excel as needed. Ideally the report should stand alone. A client may wish to see the Excel work, but should understand key points about what you did without looking at it.

2. Turn in the Excel spreadsheet anyhow, since I do need to check that your spreadsheet work is accurate.