need help on Time Series Forcasting
Hi, I'm working on my assignment about Time Series Forcasting of a book called "Predictive Analytics and Data Mining".
My assignment is to get the same result on the book, but I'm stuck in the middle of steps. The book doesn't show any specific steps to ge that result. I have to make that result just by refering to this:
"This example used Linear Regression and making inputYt5 - inputYt0 as independent variables. Make label variable as suborniate variable.Independent variables range from June 2011 to Nov 2011. Target or label variable starts from Dec 2011."
The book I have is not written in English so having trouble translating all the sentences but those are the main points that are given to me to make that result.
Please see my pdf file which I posted below.
Answers
Hello,
It looks like you just need some guidance on using the Windowing operator. The windowing operator allows you to look back a number of steps in time to create a flat single example to apply your algorithm to. For example, you can look at how the customer's previous 2 interactions affect their most recent purchase behaviour.
You set how far you want to look back with 'Window Size'. This will generate the attributes.
But next you want to know how to predict a variable from that. Tick the box 'Create Label' and a new option will appear to select which attribute you want to use all the information you have transformed into a single row to predict. Normally the next record in the series.
Thank you for your kind answer, Edward. But if you look at my pdf file, I ticked the create label and set the label attribute as Date and 1 as horizon.