如何调整随机森林代码以进行质量预测

Question

I am new to machine learning.I have got this dataset - http://archive.ics.uci.edu/ml/datasets/Wine+Quality . I have to predict Quality of wine which is the last column on the dataset.I thought about applying Neural network or random forest for this, where as NN gave around 55% accuracy, with random forest I managed to get 73% so far. I want to improve the accuracy further.Below is code I have written.

wineq <- read.csv("wine-quality.csv",header = TRUE)
str(wineq)

wineq$taste <- ifelse(wineq$quality < 6, 'bad', 'good')
wineq$taste[wineq$quality == 6] <- 'normal'
wineq$taste <- as.factor(wineq$taste)
set.seed(54321)
train <- sample(1:nrow(wineq), .75 * nrow(wineq))
wineq_train <- wineq[train, ]
wineq_test  <- wineq[-train, ]

library(randomForest)

rf=randomForest(taste~.- 
quality,data=wineq_train,importance=TRUE,ntree=100)

rf_preds = predict(rf,wineq_test)
rf_preds
table(rf_preds, wineq_test$taste)

Output:

table(rf_preds, wineq_test$taste)

rf_preds bad good normal
bad    302   11     81
good     7  163     36
normal  93  101    431

If I want to use tuneRF it gives me below error:

   fgl.res <- tuneRF(x = wineq[train, ], y= wineq[-train, ], 
   stepFactor=1.5)

Error in randomForest.default(x, y, mtry = mtryStart, ntree = ntreeTry,
: length of response must be the same as predictors

Answer 1

You need to pass to tuneRF the feature variables as x and the response variable as y .

So, first find the column position of your response variable ( taste ):

resp_pos <- which(colnames(wineq) == "taste")

Then:

fgl.res <- tuneRF(x = wineq[train, -resp_pos ], y= wineq[-train, resp_pos], 
   stepFactor=1.5)

I noticed also that you use wineq$taste <- ifelse(wineq$quality < 6, 'bad', 'good') to find your "new" response ( taste ), based on the column quality . Note that this is perfectly fine, but you need to remove the column quality before training.

If you don't do this, your model will be overly optimistic, since it'll pick up that for example:

quality < 6 will always mean taste=="bad"