multiview applies the method described in Ye & Sugihara (2016) for
forecasting, where multiple attractor reconstructions are tested, and a
single nearest neighbor is selected from each of the top k
reconstructions to produce final forecasts.
multiview(block, lib = NULL, pred = NULL, norm = 2, E = 1, tau = -1,
tp = 1, max_lag = 3, num_neighbors = "e+1", k = "sqrt", na.rm = FALSE,
target_column = 1, stats_only = TRUE, save_lagged_block = FALSE,
first_column_time = FALSE, exclusion_radius = NULL, silent = FALSE)either a vector to be used as the time series, or a data.frame or matrix where each column is a time series
a 2-column matrix, data.frame, 2-element vector or string of row indice pairs, where each pair specifies the first and last *rows* of the time series to create the library. If not specified, all available rows are used
(same format as lib), but specifying the sections of the time series to forecast. If not specified, set equal to lib
the distance measure to use. see 'Details'
the embedding dimensions to use for time delay embedding. The default value of 1 does not embed the data.
the time-delay offset to use for time delay embedding
the prediction horizon (how far ahead to forecast)
the maximum number of lags to use for variable combinations. If max_lag == 3, a variable X will be embedded with lags X[t], X[t + tau], X[t + 2*tau]
the number of nearest neighbors to use. Note that the default value will change depending on the method selected. (any of "e+1", "E+1", "e + 1", "E + 1" will set this parameter to E+1 for each run.)
the number of embeddings to use for ensemble averaging. "sqrt" or 0 will use k = sqrt(m) where m is the number of multiview combinations of the set of input variables
logical. Should missing values (including `NaN`` be omitted from the calculations?)
the name (index) of the column to forecast
specify whether to output just the forecast statistics or the raw predictions for each run
specify whether to output the lagged block that
is constructed as part of running multiview
indicates whether the first column of the given block is a time column and excluded when building the library
excludes vectors from the search space of nearest neighbors if their *time index* is within exclusion_radius (NULL turns this option off)
prevents warning messages from being printed to the R console
A named list with items "View" and "Predictions". View is a data.frame with components:
| col_i,... col_j | column indices of the embedding |
| name_i,... nam_j | column names of the embedding |
| rho | correlation of the projection |
| MAE | maximum absolute error of the projection |
| RMSE | root mean square error of the projection |
Predictions is a data.frame of the predictions from the best multivew ensemble.
multiview uses multiple time series given as input to
generate an attractor
reconstruction, and then applies the simplex projection to make
forecasts. This method generalizes the simplex routine,
and allows for "mixed" embeddings, where multiple
time series can be used as different dimensions of an attractor
reconstruction.
The default parameters are set so that, given a matrix of time series, forecasts will be produced for the first column. By default, all possible combinations of the columns are used for the attractor construction, the k = sqrt(m) heuristic will be used, forecasts will be one time step ahead. If a time vector is not supplied, 1:NROW will be used. The default lib and pred are to use the first half of the data for the "library" and to predict over the second half of the data. Unless otherwise set, the output will be just the forecast statistics.
norm = 2 (only option currently available) uses the "L2 norm",
Euclidean distance:
$$distance(a,b) := \sqrt{\sum_i{(a_i - b_i)^2}}
$$
# NOT RUN {
block <- block_3sp[, c(2, 5, 8)]
multiview( block, k=10 )
# }
Run the code above in your browser using DataLab