Impute missing values and attach lagged features
Usage
impute_all(
dt,
period,
nthreads_dt = 32L,
nthreads_collapse = 32L,
nthreads_imputation = 32L
)Arguments
- dt
The input data table to be imputed.
- period
The period for lagged features.
- nthreads_dt
The number of threads to be used for data.table operations.
- nthreads_collapse
The number of threads to be used for collapse operations.
- nthreads_imputation
The number of threads to be used for the imputation process.
Note
This function performs imputation on a given data table by replacing missing values with imputed values. It follows a series of steps including data cleaning, name cleaning, geoscf column renaming, NDVI 16-day backward filling, zero-variance exclusion, excessive "true zeros" exclusion, and imputation using missRanger. A few points should be discussed to sophisticate the imputation process: exclusion threshold for rates of zero observations, which might lead to significant improvement in the imputation process especially in terms of speed and accuracy.