In sungcheolkim78/FiDEL: Fermi-Dirac based Ensemble Learning

knitr::opts_chunk$set(echo = TRUE)
library(FiDEL)

FD Ensemble method

auclist <- create.auclist(0.51, 0.95, 50)
data_rank <- create_predictions(n=1000, auclist=auclist, p=0.5)

fde1 <- fde(data_rank$predictions)
fde1 <- calculate_performance(fde1, data_rank$actual_labels, alpha=7)

plot performance

#plot_performance(fde1, nsample=100, trendline=F)
plot_performance_nmethods(fde1, nmethod_list = 3:10, nsample=100)

check correlation

(cor_m <- cor(fde1@rank_matrix))

plot_cor(fde1, class_flag='positive')

Aggregate predictions

#fde1 <- calculate_performance(fde1, data_rank$actual_labels)
fde1 <- predict_performance(fde1, auclist, 0.1)

(cor_m <- cor(fde1@logit_matrix))

Logit distribution

plot_Logit <- function(fde_d, idx) {
  y <- fde_d@logit_matrix[, idx]
  x <- rank(y)
  plot(x, y, col='black', xlab = 'Rank', ylab='Logit')
  points(x, fde_d@logit_matrix[, 1], col='gray')
  points(x, fde_d@logit_matrix[, 2], col='blue')
  points(x, fde_d@logit_matrix[, 3], col='green')
  points(x, fde_d@logit_matrix[, 4], col='pink')
  points(x, fde_d@estimated_logit, col='red')
}

plot_Logit(fde1, 2)

beta_en <- function(k) {
  mean(fde1@rank_matrix[k,1:fde1@nmethods]*fde1@beta)/fde1@estimated_rank[k]
}

plot(sapply(1:fde1@nsamples, beta_en))

plot.scores(fde1@estimated_logit, data_rank$actual_labels)

plot.scores(fde1@estimated_logit, fde1@estimated_label)

plot_FDstatistics(fde1)

plot_ensemble(fde1, method='invauc', alpha=0.85)

plot_single(fde1, "score", c=9)

plot_single(fde1, "pcr", c=9, n=100, m=300)

plot_single(fde1, "roc", c=9, n=100, m=300)

sungcheolkim78/FiDEL documentation built on Nov. 13, 2024, 7:58 a.m.