get_time_wts: Calculate temporally-biased re-sampling weights

In mrfoliveira/STResampling-DSAA2019: Biased Resampling Strategies for Imbalanced Spatio-Temporal Forecasting

Description

Calculate weights for re-sampling with a temporal bias. Most recent observations have weights that tend to 1, while the oldest observations have weights that tend to 0 (meaning they are less likely to be kept). Most recent observations: w -> 1; oldest: w -> 0.

Usage

get_time_wts(times, phi, rel.thr)

Arguments

times	a vector of time-stamps
phi	a vector of the relevance values of df's target variable
rel.thr	a relevance threshold above which an observation is considered relevant

Value

A vector of temporally-biased re-sampling weights, scaled to fit within range [0,1].

Author(s)

Mariana Oliveira

mrfoliveira/STResampling-DSAA2019 documentation built on April 9, 2021, 5:39 a.m.