Create a time series dataset from a torch_tensor matrix

Create a time series dataset from a torch_tensor matrix

ts_dataset(
  data,
  timesteps,
  horizon,
  jump = horizon,
  predictors_spec = list(x = NULL),
  outcomes_spec = list(y = NULL),
  categorical = NULL,
  sample_frac = 1,
  scale = TRUE,
  extras = NULL,
  ...
)

Arguments

data: (torch_tensor) An input data object. For now it only accepts two-dimensional tensor, i.e. matrices. Each row is a timestep of a single time series.
timesteps: (integer) Number of timesteps for input tensor.
horizon: (integer) Forecast horizon: number of timesteps for output tensor.
jump: (integer) Jump length. By default: horizon length.
predictors_spec: (list) Input specification. It should be a list with names representing names of tensors served by dataset, and values being feature indices.
outcomes_spec: (list) Target specification. It should be a list with names representing names of tensors served b
categorical: (character) Names of specified column subsets considered as categorical. They will be provided as integer tensors.
scale: (logical or list) Scale feature columns. Boolean flag or list with mean and sd values.
extras: (list) List of extra object to be stored inside the ts_dataset object.
sample_fram: (numeric) A numeric value > 0. and <= 1 to sample a subset of data.

Note

If scale is TRUE, only the input variables are scale and not the outcome ones.

See: Is it necessary to scale the target value in addition to scaling features for regression analysis? (Cross Validated)

Examples

library(dplyr, warn.conflicts = FALSE)
library(torchts)

weather_pl_tensor <-
  weather_pl %>%
  filter(station == "TRN") %>%
  select(-station, -rr_type) %>%
  as_tensor(date)

# We obtained a matrix (i.e. tabular data in the form of 2-dimensional tensor)
dim(weather_pl_tensor)
#> [1] 7305    7

weather_pl_dataset <-
   ts_dataset(
     data = weather_pl_tensor,
     timesteps = 28,
     horizon = 7,
     predictors_spec = list(x = 2:7),
     outcomes_spec   = list(y = 1),
     scale = TRUE
   )

weather_pl_dataset$.getitem(1)
#> $x
#> torch_tensor
#> -1.3449 -1.6086 -0.3617 -0.3091 -0.2641  0.0491
#> -1.2028 -1.2505 -0.3617 -0.3091 -0.2641 -0.0979
#> -0.7637 -0.9046  1.6144  1.4789  1.0892 -0.1111
#> -0.5442 -0.5712 -0.3617 -0.3091 -0.2641 -0.0430
#> -0.5829 -0.7688 -0.3617 -0.3091 -0.2641 -0.0979
#> -0.3246 -0.2747 -0.3617 -0.3091 -0.2641 -0.2055
#> -0.0922  0.0464 -0.1401 -0.3091  0.0742 -0.0211
#> -0.2213 -0.0154  2.8149  0.7516  3.5983 -0.1023
#> -0.6862 -0.4724  1.1527  2.1759 -0.2641 -0.0738
#> -0.8412 -0.8799 -0.2509 -0.3091 -0.0949  0.0623
#> -0.8670 -0.9540 -0.1401  0.0546 -0.2641  0.0865
#> -1.0220 -0.9911 -0.2694 -0.2182 -0.2077  0.1962
#> -1.1641 -1.3122 -0.2879 -0.2485 -0.2077  0.3806
#> -1.6032 -1.7939 -0.3433 -0.2788 -0.2641  0.4618
#> -1.2932 -1.0529 -0.3617 -0.3091 -0.2641  0.4245
#> -1.4095 -1.5963 -0.3617 -0.3091 -0.2641  0.3608
#> -1.7065 -1.6704 -0.3617 -0.3091 -0.2641  0.3038
#> -1.9131 -1.9174 -0.3617 -0.3091 -0.2641  0.2884
#> -2.1198 -2.0780 -0.3617 -0.3091 -0.2641  0.2928
#> -2.1972 -2.1521 -0.3617 -0.3091 -0.2641  0.3060
#> -2.0294 -1.9545 -0.3617 -0.3091 -0.2641  0.3433
#> -1.3320 -1.0652 -0.3617 -0.3091 -0.2641  0.2401
#> -1.2157 -1.0652 -0.3617 -0.3091 -0.2641  0.0974
#> -0.8800 -0.8799 -0.0662 -0.3091  0.1870  0.0338
#> -0.3892 -0.3118  1.1342 -0.2788  1.9913  0.0030
#> -0.3634 -0.4353  0.3401 -0.3091  0.8072 -0.1001
#> -0.4538 -0.4724 -0.3617 -0.3091 -0.2641 -0.2340
#> -0.6733 -0.7688 -0.3617 -0.3091 -0.2641 -0.1813
#> [ CPUFloatType{28,6} ]
#> 
#> $y
#> torch_tensor
#> -0.8771
#> -1.3207
#> -1.4820
#> -1.5223
#> -1.4619
#> -1.7542
#> -1.8046
#> [ CPUFloatType{7,1} ]
#>