Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

EMET8012 Computer Lab 6

Report Due: 27 Oct 2023

In this optional project, we will consider the highway traffic data used in Part III of Peixeiro’s textbook.  The data file can be downloaded from the course Wattle website or Peixeiro’s github.  If you download data from the github, please make sure that you download the “preprocessed” data. The raw data have all sorts of weird problems such as two

observations for the same hour due to a change of weather during that hour.

We aim to forecast traffic volumes in the next 24 hours.  In other words, the width of the forecasting window is 24 hours. We use the first 70% of observations as the training set, the next 20% of observations as the validation set and the

last 10% of observations as the test set.

The benchmark forecast used in this project is a naive seasonal forecast: use the observation from the previous day

as the forecast. The performance measure is mean absolute errors (MAE).

Apart from the performance of your model, you will also be marked on your general design, your model selection

approach and your analysis of the results. Please note the following rules.

1.  You must consider at least two forecasting models not including the benchmark forecast.

2.  If we find that you use the test set in your model building and model selection process, you will receive at most 30/100 marks from this lab.  Using the test set in model selection is one of the most serious errors in forecasting and considered cheating in this project.

3.  If your forecasting model cannot beat the benchmark forecast, you will receive at most 60/100 marks from this project.

4.  I find results reported in Peixeiro’s book dubious. If you decide to use some of his code, do this with care.

Due to the time constraint on this project, there is no hurdle task.

Good luck!