Bike Sharing Dataset - Random Forest basic prediction with validation of variables used at training time

bikesharing.R

# I got this working with code from https://datascienceplus.com/random-forests-in-r/
# It is an excelent resource!
set.seed(50)
bikeperday <- read.csv('Bike-Sharing-Dataset/day.csv')

# Calculate sample
sample <- sample.int(n = nrow(bikeperday),
                     size = floor(.75 * nrow(bikeperday)),
                     replace = F)

train <- bikeperday[sample, ]
test <- bikeperday[-sample, ]

# Let's use random forests to create a basic prediction
require(randomForest)

# First, we try to decide the number of variables used at train time
oob.err=double(7) # Out of Bag error, i.e. errors at train time
test.err=double(7) # Test error
for(tries in 1:7) {
  model <- randomForest(cnt ~ atemp * hum * windspeed * season * mnth * weathersit * workingday,
                        data = train, mtry = tries, ntree = 800)
  oob.err[tries] <- model$mse[800] # We get the mean squared error of the last tree
  
  pred <- predict(model, test)
  test.err[tries] <- with(test, mean((cnt - pred)^2)) # Mean squared Error of predicted value
  
  varImpPlot(model)
  cat(paste0('Try number', tries), ' ') # Show in console
}

# Plot the errors
matplot(1:7, cbind(oob.err,test.err), pch=19,
                                      col=c("red","blue"),
                                      type="b",
                                      ylab="Mean Squared Error",
                                      xlab="Number of Predictors Considered at each Split")

legend("topright", legend=c("Out of Bag Error","Test Error"), pch=19, col=c("red","blue"))

# Based on the results, let's use 3 variables on the model training
final_model <- randomForest(cnt ~ atemp * hum * windspeed * season * mnth * weathersit * workingday,
                            data = train,
                            mtry = 3,
                            ntree = 500)

varImpPlot(final_model)

# Let's do some prediction
newdata <- data.frame(atemp = 35.3, # Celcius
                      season = 2, # summer
                      hum = 0.8, # 80 %
                      mnth = 6, # June
                      windspeed = 0, # No wind. Values up to x/67
                      weathersit = 1, # Clear skies
                      workingday = 1)

pred <- predict(final_model, newdata)
print(paste('Approx Number of rented bikes: ', pred))