My two goals are: first to have a new forecast method that is better than the current forecast method that I have and second to learn how to run the forecast that you generate. Most likely SARIMA will improve the WAPE but I'm open to any method as long as it is more accurate.
Attached is a sample data with Actuals for 2014, 2015, 2016, 2017 and the forecast I generated for 2017 with it's corresponding WAPE. I'd like to be able to backtest the forecast for forecasts generated on Week -1, Week-2, Week-3 and Week-4.
For each week you would generate four forecasts. As an example let's say we are forecasting year week 2017 10.
* Week -1 is the forecast you generate on year week 2017 09 and you can only see data from week 2017 08 and before.
* Week -2 is generated on year week 2017 08 and you can only use data from week 2017 07 and before.
* Week -3 is the forecast you generate on year week 2017 07 and you can only see data from week 2017 06 and before.
* Week -4 is generated on year week 2017 06 and you can only use data from week 2017 05 and before.
The WAPE calculation that needs to be improved is Absolute(Actuals-Forecast)/Forecast.
14 pekerja bebas membida secara purata $143 untuk pekerjaan ini
Hi... I am a statistics specialist, certified by Freelancer. I fully understand your project and I am sure I can help you. Let's discuss details by chat.
Hi, We could use SARIMA and hybrid forecasting model using R programming. I did similar projects before. After testing several model i could recommend the most accurate one. Regards Ismail
"Quality is Not an Act it is a Habit" I believe that my experience and skill in this background will prove to be of great help to you. Contact me to discuss more on the detail
Already working in the same field, Anticipating the forecast for a well established firm on PAN INDIA level. MIS Reports on day to day basis, Margin Working, etc.