Narrowed question below
I'm researching forecasting models for my job. I need a forecasting model for which I made the following assumpions:
- I need a Time series model
- My data is influenced by "seasons" (There are four months in which the data is higher, the months are spread throughout the year, it's month 4, 6, 9 and 10)
- I need an accurate forecast between 1 and 6 months.
- I guess it's stationary but I'm not sure
- I have accurate data of the last 3 years
I found the models below, the problem is, I can't seem to find literature that lets me choose between the models. My boss wants me to have chosen a model before I'm going to test it on my data. With every bit of literature I seek and find, all the models are getting more unclear and I'm stuck at this point for a while now.
The models I found:
- Seasonal ARIMA
- Unobserved Component Model
- Seasonal Naïve Method
- Method of Holt-Winters
- Box-Jenkins (Is this the same as seasonal ARIMA?)
- STL Decomposition
I'm hoping you're able to help me clear up which models are used for what, how to make a choice between the models, if some models aren't to be used for my data and if I forgot a model that could be the solution for me.
Edit, narrowing down my question:
I have got monthly data, there's 'spreaded seasonality', how I like to call it, four months in the year with higher numbers. Except for Seasonal Naïve Method I can't seem to find a forecasting model which is applicable to this kind of seasonality. I've looked at lots of different models and literature but this doesn't get me much further. The data itself is almost constant throughout the year, except for those four months. I need to forecast up to 6 months from when the model is used.
Is there a forecasting model that can be used for accurately forecasting data with this 'spreaded seasonality'?