R - caret data handling

caret.R

library(caret)

#looking at missing value
options(digits=2)
stats <- data.frame(missing = sapply(iris, function(x) sum(is.na(x))),
           mean = sapply(iris, function(x) if(is.numeric(x)) {mean(x, na.rm = T)} else names(table(x)[order(table(x), decreasing =  T)])[1])
)



#Generate Dummy Variable
head(predict(dummyVars(Sepal.Length ~ ., data = iris), newdata= iris))

#nearZeroVar
nzv <- nearZeroVar(iris, freqCut = 50/51, saveMetrics= TRUE)
nzv[nzv$nzv,][1:5,]

#find Correlation
cor.index <- findCorrelation(cor(iris[,-5]), cutoff = .75)
names(iris[,cor.index])

#find Linear combo
linear.index <- findLinearCombos(iris[,-5])
names(iris[, linear.index])

#
#preProcess
pre.value <- preProcess(iris[1:100,], method = c("center", "scale"))
iris.train.trans <- predict(pre.value, iris[1:100,])
iris.test.trans <- predict(pre.value, iris[101:150,])

#age group annomally value
assign to normally threshold

#assign missing value by group
new.incomes <-df %>%
  select(nomprov) %>%
  merge(df %>%
          group_by(nomprov) %>%
          summarise(med.income=median(renta,na.rm=TRUE)),by="nomprov") %>%
  select(nomprov,med.income) %>%
  arrange(nomprov)
df <- arrange(df,nomprov)
df$renta[is.na(df$renta)] <- new.incomes$med.income[is.na(df$renta)]
rm(new.incomes)

df$renta[is.na(df$renta)] <- median(df$renta,na.rm=TRUE)
df <- arrange(df,fecha_dato)

#print out unique values
char.cols <- names(iris)[sapply(iris,is.character)]
for (name in char.cols){
  print(sprintf("Unique values for %s:", name))
  print(unique(iris[[name]]))
  cat('\n')
}

#Convert to binary then apply below
head(diamonds)
a <- predict(dummyVars(~ color + cut, data = diamonds, fullRank = T, drop2nd = T), newdata= diamonds)
head(sapply(a, function(x) if (x>0) {1} else {0}))
dummies <- predict(dummyVars(~ Species, data = iris), newdata = iris)
head(dummies, n = 10)

dummies <- predict(dummyVars(~., data = diamonds), newdata = diamonds)
head(dummies, n = 10)



#A
totals.by.feature <- df %>%
  group_by(month,feature) %>%
  summarise(counts=n())

https://www.kaggle.com/apryor6/santander-product-recommendation/detailed-cleaning-visualization

df %>% 
  group_by(month,feature,status) %>%
  summarise(counts=n())%>%
  ungroup() %>%
  inner_join(totals.by.feature,by=c("month","feature")) %>%
  
  mutate(counts=counts.x/counts.y) %>%
  ggplot(aes(y=counts,x=factor(month.abb[month],levels=month.abb[seq(12,1,-1)]))) +
  geom_bar(aes(fill=status), stat="identity") +
  facet_wrap(facets=~feature,ncol = 6) +
  coord_flip() +
  my_theme_dark + 
  ylab("Count") +
  xlab("") + 
  ylim(limits=c(0,1)) +
  ggtitle("Relative Service \nChanges by Month") +
  theme(axis.text   = element_text(size=10),
        legend.text = element_text(size=14),
        legend.title= element_blank()      ,
        strip.text  = element_text(face="bold")) +
  scale_fill_manual(values=c("cyan","magenta"))