Evaluate Orange Time Series Models.

Assignment: Time Series

Overview

This assignment asks you to use forecasting tools to predict the price of gas based on the historical record of prices in Ottawa.

Part 1: Collecting the Data

Download the Fuel Assignment Source Data.csv file from Moodle (see Course Data Files – Chapter 7) and save it as an Excel file. Manipulate the data to isolate only the data that we are interested in – the price of Regular Unleaded Gasoline in Ottawa, on or after June 6, 2005. You can separate this data by either editing it directly in Excel, or by utilizing data processing tools in Orange. Doing it in Excel relies on simple skills you have from 1150, utilizing the data processing tools in Orange is better practice for using the Orange tool going forward; pick whichever you are comfortable with. This new sheet is what you will use for your forecast.

Note: Replace the date value with an index for the week (1,2,3,4,5…), this will make things much more simple for you.

Part 2: Evaluate Orange Time Series Models

Create a workflow in Orange that imports the isolated data, converts it to a time series, and evaluates the accuracies of the VAR Model and ARIMA Model against each other.

Part 2A: Default Settings

Compare the accuracies of the two models via the model evaluation object utilizing the default settings. Note the results for your records.

Note: If not already so, set the confidence interval for both models at 80.

Part 2B: Manipulate Options

For both the ARIMA and VAR models, change some settings in an attempt to make predictions more accurate. Try settings of:

If you are feeling really statistical you can further manipulate these options to attempt to minimize your errors and create the most accurate prediction model. Only the specified results will be for marks, but it is good practice for the more involved exercises later in the course.

Part 4: Produce Forecasts

Generate forecasts for the next 100 weeks for each model. (Ensure that if you have multiple models in your workflow – such as one for evaluation and one for forecasting – that they utilize the same settings.)

Grading

Deliverables

The deliverables for this assignment are:

• Answers to the question set posted on Moodle. Each person must individually answer the quiz!
• A screenshot of your Orange workspace submitted in the Moodle dropbox. One per person.

Note: The answers to all the Moodle questions can be pulled directly from your results once you’ve completed the above parts.