I’m working on a ML project. I’m trying to impute a few values of a time series of Afghanistan. To be more specific, the series representes the percentage of people that has access to electricity.

This is the code I’m using:

library(fable)
library(tsibble)

data3 = data2%>% as_tsibble(index = Year ,key = Country)

afg<-data3[data3['Country']=='Afghanistan',]

df_impute= afg%>%
  model(TSLM(V1 ~ trend())) %>% 
  interpolate(afg)

It does work with no issues at all, but the problem is that in some cases i get absurd values, for example:

# A tsibble: 6 x 3 [1Y]
# Key:       Country [1]
  Country      Year    V1
  <fct>       <int> <dbl>
1 Afghanistan  1960 -195.
2 Afghanistan  1961 -190.
3 Afghanistan  1962 -185.
4 Afghanistan  1963 -180.
5 Afghanistan  1964 -175.
6 Afghanistan  1965 -170.

or like the one for 2022:

# A tsibble: 6 x 3 [1Y]
# Key:       Country [1]
  Country      Year    V1
  <fct>       <int> <dbl>
1 Afghanistan  2017  97.7
2 Afghanistan  2018  93.4
3 Afghanistan  2019  97.7
4 Afghanistan  2020  97.7
5 Afghanistan  2021  97.7
6 Afghanistan  2022 113. 

The only solution I found is this one:

df_impute= afg%>%
  model(TSLM(V1 ~ trend())) %>% 
  interpolate(afg)%>%
    mutate(V1 = pmax(pmin(V1, 100), 0))

but I don’t think is the right solution, because I feel like this is a wrong way to impute.
Is there a way to specify the ranges?