tsensembler (version 0.0.5)

constructive_aggregation: Constructive aggregation constructor

Description

Constructive aggregation via out-performance contiguity This method denotes the idea of rearranging a portfolio of models (base_ensemble) into different overlapping subsets. These subsets are aggregated (aggregate_subsets) into combined opinions, forming new models. These models are combined into a final decision through aggregate_hypos.

Usage

constructive_aggregation(form, data, specs, lambda = 100, alpha = 30,
  depth = NULL, aggregate_subsets = "simple",
  aggregate_hypos = "simple")

Arguments

form

formula

data

training data

specs

object of class model_specs-class. Contains the parameter setting information for training the base models;

lambda

smoothing window size

alpha

contiguity interval size

depth

depth size how large is the maximum size of the subsets. If NULL, defaults to no. of predictors minus one.

aggregate_subsets

aggregation approach for the set of subsets.

aggregate_hypos

final aggregation approach. How should the combined opinions be aggregated.

Examples

# NOT RUN {
specs <- model_specs(
  learner = c("bm_svr", "bm_mars"),
  learner_pars = list(
    bm_glm = list(alpha = c(0, .5, 1)),
    bm_svr = list(kernel = c("rbfdot"),
                  C = c(1, 3))
  )
)

data("water_consumption")
waterc <- embed_timeseries(water_consumption, 5)
train <- waterc[1:300, ] # toy size for checks
test  <- waterc[301:320, ] # toy size for checks

model <- constructive_aggregation(target ~., train, specs, 10,5,NULL,"window_loss","simple")
preds <- predict(model, test)

# }