A representation as a block of predictors

Lays a [unit, bin, channel] array out as one row per unit and one column per bin of each channel, which is what a model reading a table of predictors takes. It is what the learners that read a tabular block fit on, so a learner of your own that calls it reads the same columns they do.

Usage

# S3 method for class 'timesift_matrix'
as.matrix(x, ...)

Arguments

x

A grain_matrix(), lookback_matrix(), feature_matrix() or build_representation() result.

...

Ignored.

Value

A numeric [unit, predictor] matrix, the units as row names.

Details

Columns are named channel@bin, channel by channel and bin by bin within each. A channel that holds the same number in every bin, as a static predictor does, is one column named by the channel alone.

Examples

t <- seq(as.POSIXct("2021-09-01", tz = "UTC"), by = "hour", length.out = 24 * 60)
d <- data.frame(plot = rep(c("a", "b"), each = length(t)), t = rep(t, 2),
                temp = sin(seq_len(2 * length(t)) / 24))
x <- grain_matrix(d, plot, t, temp, grain = "month", stats = c("min", "max"))
as.matrix(x)