I have a question regarding Holt Winters method.

I have an order history which contains at least the customers and the belonging time when the order was taken. We will skip the rest for now since it is not important to understand the problem.

What I want to do is, I want to predict when the customer will order next. So, I'm not interested in the amount of a certain product, just when will the customer order next. I have a real-life data set (so no pre-optimized stationary data set).

I thought its a good idea to use a time series. When I will plot the data visually, the x-axis will contain the time and the y-axis consists of 0 and 1 values, where 1 denotes the customer brought something and 0 he doesn’t.

What I want to understand is, if the Holt Winters approach is capable of dealing with zero/one values well? I found this article which covers my case but I'm not sure if the article uses that case just for explanation purposes or if it would work in practise as well. What they are doing is, they map the zero/one values relative to the average. I guess its because of preventing zero/one combinations.

It would be great if someone can share his/her experience with that.

Thanks Raphael

I have a question regarding Holt Winters method.

I have an order history which contains at least the customers and the belonging time when the order was taken. We will skip the rest for now since it is not important to understand the problem.

What I want to do is, I want to predict when the customer will order next. So, I'm not interested in the amount of a certain product, just when will the customer order next. I have a real-life data set (so no pre-optimized stationary data set).

I thought its a good idea to use a time series. When I will plot the data visually, the x-axis will contain the time and the y-axis consists of 0 and 1 values, where 1 denotes the customer brought something and 0 he doesn’t.

What I want to understand is, if the Holt Winters approach is capable of dealing with zero/one values well? I found this article which covers my case but I'm not sure if the article uses that case just for explanation purposes or if it would work in practise as well. What they are doing is, they map the zero/one values relative to the average. I guess its because of preventing zero/one combinations.

It would be great if someone can share his/her experience with that.