backlog/testing.R

In mayer79/outRanger: Multivariate Outlier Detection and Replacement

#=====================================================================================
#  TESTING THE PACKAGE "outForest"
#=====================================================================================

# library(outForest)
library(ranger)
library(missRanger)
library(FNN)
lapply(list.files("r", full.names = TRUE), source)
out <- outForest(as.data.frame(ggplot2::diamonds), num.trees = 10, max_n_outliers = 10)
outliers(out)
head(Data(out))

out <- outForest(CO2, allow_predictions = TRUE)
summary(out)
head(CO2)
X <- head(CO2)
X[1, "uptake"] = 40
outliers(predict(out, X))

# F) Time series data
set.seed(3)
t <- seq(0, 10, by = 0.01)
y <- sin(t) + runif(length(t), 0, 0.3)
y[500] <- 0
plot(y ~ t, type = "l")
ch <- outForest(data.frame(t, y), min.node.size = 40)
outliers(ch)

# G) Bathtube
set.seed(3)
t <- seq(0, pi, by = 0.01)
dat <- transform(data.frame(t), x = cos(t), y = sin(t) + runif(length(t), -0.1, 0.1))
dat[150, 2:3] <- c(0, 0.4)
plot(y ~ x, data = dat, cex.lab = 1.3, cex.axis = 1.3)
ch <- outForest(dat[-1], allow_predictions = TRUE)
outliers(ch)

hd <- head(dat)
hd[5, "y"] <- 3
pred <- predict(ch, hd)
summary(pred)

# Testing NAs
dat <- generateNA(as.data.frame(ggplot2::diamonds))
out <- outForest(dat, num.trees = 10, max_n_outliers = 10)
outliers(out)
head(Data(out))


# G) Bathtube
set.seed(3)
t <- seq(0, pi, by = 0.01)
dat <- transform(data.frame(t), x = cos(t), y = sin(t) + runif(length(t), -0.1, 0.1))
dat <- generateNA(dat, p = c(y = 0.05))
dat[150, 2:3] <- c(0, 0.4)
head(dat)
plot(y ~ x, data = dat, cex.lab = 1.3, cex.axis = 1.3)
ch <- outForest(dat[-1], allow_predictions = TRUE)
head(Data(ch))
outliers(ch)

hd <- head(dat)
hd[5, "y"] <- 3
hd[3, "x"] <- 3
hd[3, "y"] <- NA
hd
pred <- predict(ch, hd)
summary(pred)
Data(pred)