Empirically extrapolates recurring patterns in numeric and categorical time-feature sequences. Candidate windows are selected by similarity, validated with rolling-origin evaluation, and summarized as forecast distributions. The runtime package uses only base R.
naive performs dependency-free empirical extrapolation of numeric and
categorical time-feature sequences.
fit <- naive::naive_fit(data.frame(x = sin(seq(0, 10, length.out = 100))),
seq_len = 8, n_windows = 3, n_samp = 5)
print(fit)
plot(fit)
Version 2.0.0 uses only base R at runtime. Tests and the vignette use optional
development packages listed in Suggests.