optimal_stage {dCUR}

optimal_stage

Description

optimal_stage is a function used to select the optimal k, the number of columns and rows of dynamic CUR object; it also produces a data frame and corresponding plots.

Usage

optimal_stage(data, limit = 80)

Arguments

data

An object resulting from a call to dCUR.

limit

Cumulative percentage average of relative error rate.

Details

Select the optimal stage of dynamic CUR descomposition

The objective of CUR decomposition is to find the most relevant variables and observations within a data matrix to reduce the dimensionality. It is well known that as more columns (variables) and rows are selected, the relative error will decrease; however, this is not true for k (number of components to compute leverages). Given the above, this function seeks to find the best-balanced stage of k, the number of relevant columns, and rows that have an error very close to the minimum, but at the same time maintain the low-rank fit of the data matrix.

Value

data

a data frame which specifies the relative error for each stage of CUR decomposition.

rows_plot

a plot where the average relative error is shown for each number of relevant rows selected.

columns_plot

a plot where the average relative error is shown for each number of relevant columns selected.

k_plot

a plot where the average relative error is shown for each k (number of components to compute leverage), given the optimal number of relevant columns and rows.

optimal

a data frame where the average relative error is shown for optimal k (number of components to compute leverage), given the optimal number of relevant columns and rows.

Author(s)

Cesar Gamboa-Sanabria, Stefany Matarrita-Munoz, Katherine Barquero-Mejias, Greibin Villegas-Barahona, Mercedes Sanchez-Barba and Maria Purificacion Galindo-Villardon.

References

Villegas G, Martin-Barreiro C, Gonzalez-Garcia N, Hernandez-Gonzalez S, Sanchez-Barba M, Galindo-Villardon M (2019). “Dynamic CUR, an alternative to variable selection in CUR decomposition.” Revistas Investigacion Operacional, 40(3), 391–399.

See Also

dCUR CUR

Examples

results <- dCUR(data=AASP, variables=hoessem:notabachillerato,
k=15, rows=0.25, columns=0.25,skip = 0.1, standardize=TRUE,
cur_method="sample_cur",
parallelize =TRUE, dynamic_columns  = TRUE,
dynamic_rows  = TRUE)
result <- optimal_stage(results, limit = 80)
result
result$k_plot
result$columns_plot
result$data
result$optimal