search

kaushikrch / rucm

Implementation of Unobserved Components Models (UCM) in R
3 stars 3 forks source link

UCM Forecast resulting 0 MAPE in Train period #15

Open akhilbhat060890 opened 3 years ago

akhilbhat060890 commented 3 years ago

Hi Team,

I am facing a very weird issue where the forecast values for train period are coming exactly same as the input variables although the test period is reasonably ok. What could be the reason behind the same.

I am using below:

ucm_formula = ucm_formula = as.formula("dependent ~ cluster2")
ucm_formula

#UCM model
UCM_fit <- ucm(formula =  ucm_formula,                                                                                                                                                                                  
               data = train,
               #slope =TRUE,
               #season = TRUE,
               level = TRUE,
               #season.var = 0,
               #slope.var =0,
               #level.var = 0,
               #irregular.var = 0,
               #season.length = 52
)
UCM_fit

cluster2 is a cluster of variables. I have been running this for other datasets and worked fine but not sure why this is happening. i have tried playing around with parameters within UCM but it is still giving MAPE of 0 for train dataset.

Can you please help with the same. Thanks