Notebook Abalone - K-nearest neighbor (KNN)

What KNN is about

Non-parametric method used for both classification and regression.

Most of the times it is used for classification problems, where the response variable is categorical

Data Preparation

abalone <- read.csv("abalone.csv", header =T, na.strings=c("","NA"))
suppressWarnings(suppressMessages(library(dplyr)))
#create an sex column that is numeric
abalone <- abalone %>%
  mutate(sex_num =case_when(
    sex %in% 'M' ~ 0,
    sex %in% "F"  ~ 1,
    sex %in% "I" ~ 2
  ))
#create an age column
abalone <- abalone %>%
  mutate(age=case_when(
    rings %in% 1:5 ~ "young",
    rings %in% 6:13 ~ "adult",
    rings %in% 14:30 ~ "old"
  ))
# remove rings, sex
abalone <- abalone[c(-1, -9)]
str(abalone)

'data.frame':   4177 obs. of  9 variables:
 $ length  : num  0.455 0.35 0.53 0.44 0.33 0.425 0.53 0.545 0.475 0.55 ...
 $ diameter: num  0.365 0.265 0.42 0.365 0.255 0.3 0.415 0.425 0.37 0.44 ...
 $ height  : num  0.095 0.09 0.135 0.125 0.08 0.095 0.15 0.125 0.125 0.15 ...
 $ weight  : num  0.514 0.226 0.677 0.516 0.205 ...
 $ shucked : num  0.2245 0.0995 0.2565 0.2155 0.0895 ...
 $ viscera : num  0.101 0.0485 0.1415 0.114 0.0395 ...
 $ shell   : num  0.15 0.07 0.21 0.155 0.055 0.12 0.33 0.26 0.165 0.32 ...
 $ sex_num : num  0 0 1 0 2 2 1 1 0 1 ...
 $ age     : chr  "old" "adult" "adult" "adult" ...

Scale and prepare training and test set

### the dependent variable is age , with the different values young adult old
### standardize the predictors
set.seed(100)
abalone_scale <- data.frame(scale(abalone[1:8]))
### add the target variable to the data set abalone_scale 
abalone$age <- as.factor(abalone$age)
abalone_scale  <- cbind(abalone_scale, age = abalone$age)
i <- sample(4177, 2088)
abalone_train <- abalone_scale[i,]
abalone_test <- abalone_scale[-i,]

The value of K is important in the KNN algorithm, because the prediction accuracy in the test set depends on it. The optimal value of K is the value that leads to the highest prediction accuracy.

### we use the tune.knn function in the e1071 package to determine a good K number
### this function performs a 10-fold cross-validation
suppressWarnings(suppressMessages(library(e1071)))
t_knn <- tune.knn(abalone_train[,-9], factor(abalone_train[,9]), k = 1:100)
t_knn # names(t_knn) to see the list of variables

Parameter tuning of ‘knn.wrapper’:

- sampling method: 10-fold cross validation 

- best parameters:
  k
 17

- best performance: 0.1379325 

plot(t_knn)

Run the prediction

library(class)
age <- abalone_train$age
pred <- knn(train = abalone_train[,-9], test = abalone_test[,-9], cl = age, k = t_knn$best.parameters)

check the result

### get the prediction accuracy in the test set
mean(pred == abalone_test$age)

[1] 0.8721876

table(pred,abalone_test$age)

       
pred    adult  old young
  adult  1726  183    55
  old      18   41     0
  young    11    0    55

This library also shows same answers above. If you want mse, convert the above age factors into numeric.

suppressWarnings(suppressMessages(library(Metrics)))
# tp / tp+tn
accuracy(abalone_test$age, pred)

[1] 0.8721876

# classification error = 1-accuracy above
ce(abalone_test$age, pred)

[1] 0.1278124