BABS 502 – Forecasting and Time Series Prediction
Final project:
Forecasting Energy use for the Vancouver International Airport (YVR)
Student name:
Student number:
Introduction
What to include in the introduction: objective of the analysis, background information
Exploratory analysis
Create plots – what patterns do you see? What could be causing them?
Do you want to use any adjustments or transformations? Why or why not?
Developing Models (methods and results)
What types of models will you be using? Why?
Divide the data into a training set and a test set. What are the time periods for each of these?
Basic methods
Use the basic methods we have learned to develop forecasts for the test set
- mean method
- drift method
- naïve method
- seasonal naïve method
Plot the training set and test set data. Include the forecasts of the test set for each of the basic forecasting methods in a different color with a legend to explain.
Calculate the accuracy measures (RMSE, MAE, MAPE, MASE) to show how well the model forecasts for the test set. Present these in a table.
Which of these will provide the best benchmark to compare other models to? Why?
Exponential smoothing/ETS model
1. What is the model? Explain the type of model and any parameters.
|
|
2. Include a time plot of the data in black with a gap between the training set and test set data.
Show the fitted values of the model graphed in blue. Show the forecasts for the test set with a bold blue line.
|
|
3. Explain why this model is appropriate based on the features of the data
|
|
4. Quantify and discuss the goodness of fit of the model to the training set | 5. Calculate the accuracy measures (RMSE, MAE, MAPE, MASE) to show how well the model forecasts for the test set
|
6. a) Time plot of residuals | |
6. b) ACF plot of residuals | |
6. c) Ljung-Box test and/or Box-Pierce test of autocorrelations of residuals | 6. d) Histogram of residuals
6. e) Mean of residuals |
6. f) What properties do the residuals have? What information can you tell about your model from the residual diagnostics?
|
ARIMA model
1. What is the model? Explain the type of model and any parameters.
|
|
2. Include a time plot of the data in black with a gap between the training set and test set data.
Show the fitted values of the model graphed in blue. Show the forecasts for the test set with a bold blue line.
|
|
3. Explain why this model is appropriate based on the features of the data
|
|
4. Quantify and discuss the goodness of fit of the model to the training set | 5. Calculate the accuracy measures (RMSE, MAE, MAPE, MASE) to show how well the model forecasts for the test set
|
6. a) Time plot of residuals | |
6. b) ACF plot of residuals | |
6. c) Ljung-Box test and/or Box-Pierce test of autocorrelations of residuals | 6. d) Histogram of residuals
6. e) Mean of residuals |
6. f) What properties do the residuals have? What information can you tell about your model from the residual diagnostics?
|
Comparison of models
Compare your exponential smoothing/ETS model, and your ARIMA model to the basic methods. Choose a final model, and give the point forecasts for the next three years (January 2011 through December 2013) in a table. Plot the entire dataset with the forecasts from your chosen method.
Discuss any limitations of this final model, and recommendations you have to improve it.
Describe three other possible models to try (you do not have to create these models).
Model Ideas | Explain why this model is appropriate based on the features of the data |
Idea 1 | |
Idea 2 | |
Idea 3 |
Judgmental forecasting
Word limit for this section: 500 words
Conclusion
References