Automated Machine Learning for deep time series models

See also https://techtonique.github.io/nnetsauce/

LazyDeepMTS(
  verbose = 0,
  ignore_warnings = TRUE,
  custom_metric = NULL,
  predictions = FALSE,
  random_state = 42L,
  estimators = "all",
  preprocess = FALSE,
  show_progress = TRUE,
  n_layers = 3L,
  ...
)

Arguments

verbose: monitor progress (0, default, is false and 1 is true)
ignore_warnings: print trace when model fitting failed
custom_metric: defining a custom metric (default is NULL)
predictions: obtain predictions (default is FALSE)
random_state: reproducibility seed
estimators: specify regressors to be adjusted (default is 'all')
preprocess: preprocessing input covariates (default is FALSE FALSE)
n_layers: number of layers for the deep model
...: additional parameters to be passed to nnetsauce::CustomRegressor

Value

a list that you can $fit

Examples


set.seed(123)
X <- matrix(rnorm(300), 100, 3)

(index_train <- base::sample.int(n = nrow(X),
                                 size = floor(0.8*nrow(X)),
                                 replace = FALSE))
#>  [1] 60 62 80  9 11 43 30 61 74 55 93 34 57 38 84 63 46 78 70 18  2 72 75 21 88
#> [26] 71 79 59 37 91  6 40 82 28 32 49 35 67 47 20  1  5 96 99 90 98 48 51 10 53
#> [51] 65  7 87 44 36 23 89 24  4 29 45 58 33  3 54 94 76 83 66  8 26 56 14 25 13
#> [76] 50 31 92 22 16
X_train <- data.frame(X[index_train, ])
X_test <- data.frame(X[-index_train, ])

obj <- LazyDeepMTS()

res <- obj$fit(X_train, X_test)
print(res[[1]])
#>  [1] 0.9576618 0.9707051 0.9709615 0.9748545 0.9791164 0.9907240 0.9907240
#>  [8] 0.9907240 0.9907240 0.9966559 1.0095353 1.0110766 1.0142847 1.0348505
#> [15] 1.0648919 1.0672458 1.0735442 1.2412343 1.3172612 1.3286419 1.3619977
#> [22] 1.3765455 1.4272372 1.5753650 1.6729886 1.8409357 1.8473388 1.8900647
#> [29] 1.8900647