# tree = 10000
oob.err14 = numeric(10)
for (mtry in 1:10) {
  fit14 = randomForest(lable ~ .-user_id -product_id, data = train_2, mtry = mtry, ntree = 10000)
  oob.err14[mtry] = fit14$err.rate[500]
  cat('We are performing iteration', mtry, '\n')
}

plot(1:10, oob.err14[1:10], pch = 16, type = 'b',
     xlab = 'Variance Cosidered at Each Split',
     ylab = 'OOB Mean Squared Error',
     main = '4_Random Forest OOB Error Rates\nby # of Variables')
# variance important:
varImpPlot(fit14)
which(oob.err14 == min(oob.err14))
oob.err14[10]

##### best model ######
# mtry = 3
fitt0 <- randomForest(lable ~ .-user_id -product_id, data = train_2, mtry = 3, ntree = 10000)
fitt0
varImpPlot(fitt0)