The liver
package contains a collection of helper functions that make various techniques from data science more user-friendly for non-experts.
Here is an example to show how to use the functionality of the package by using the churn dataset which is available in the package.
data( churn )
str( churn )
'data.frame': 5000 obs. of 20 variables:
$ state : Factor w/ 51 levels "AK","AL","AR",..: 17 36 32 36 37 2 20 25 19 50 ...
$ area.code : Factor w/ 3 levels "area_code_408",..: 2 2 2 1 2 3 3 2 1 2 ...
$ account.length: int 128 107 137 84 75 118 121 147 117 141 ...
$ voice.plan : Factor w/ 2 levels "yes","no": 1 1 2 2 2 2 1 2 2 1 ...
$ voice.messages: int 25 26 0 0 0 0 24 0 0 37 ...
$ intl.plan : Factor w/ 2 levels "yes","no": 2 2 2 1 1 1 2 1 2 1 ...
$ intl.mins : num 10 13.7 12.2 6.6 10.1 6.3 7.5 7.1 8.7 11.2 ...
$ intl.calls : int 3 3 5 7 3 6 7 6 4 5 ...
$ intl.charge : num 2.7 3.7 3.29 1.78 2.73 1.7 2.03 1.92 2.35 3.02 ...
$ day.mins : num 265 162 243 299 167 ...
$ day.calls : int 110 123 114 71 113 98 88 79 97 84 ...
$ day.charge : num 45.1 27.5 41.4 50.9 28.3 ...
$ eve.mins : num 197.4 195.5 121.2 61.9 148.3 ...
$ eve.calls : int 99 103 110 88 122 101 108 94 80 111 ...
$ eve.charge : num 16.78 16.62 10.3 5.26 12.61 ...
$ night.mins : num 245 254 163 197 187 ...
$ night.calls : int 91 103 104 89 121 118 118 96 90 97 ...
$ night.charge : num 11.01 11.45 7.32 8.86 8.41 ...
$ customer.calls: int 1 1 0 2 3 0 3 0 1 0 ...
$ churn : Factor w/ 2 levels "yes","no": 2 2 2 2 2 2 2 2 2 2 ...
It shows that the ‘churn’ dataset as a data.frame
has 20 variables and 5000 observations.
We partition the churn dataset randomly into two groups: train set (80%) and test set (20%). Here, we use the partition
function from the liver package:
set.seed( 5 )
= partition( data = churn, prob = c( 0.8, 0.2 ) )
data_sets
= data_sets $ part1
train_set = data_sets $ part2
test_set
= test_set $ churn actual_test
The churn dataset has 19 predictors along with the target variable churn
. Here we use the following predictors:
account.length
, voice.plan
, voice.messages
, intl.plan
, intl.mins
, day.mins
, eve.mins
, night.mins
, and customer.calls
.
First, based on the above predictors, find the k-nearest neighbor for the test set, based on the training dataset, for the k = 8 as follows
= churn ~ account.length + voice.plan + voice.messages + intl.plan + intl.mins +
formula + eve.mins + night.mins + customer.calls
day.mins
= kNN( formula, train = train_set, test = test_set, k = 8 ) predict_knn
To report Confusion Matrix:
conf.mat( predict_knn, actual_test )
Actual
Predict yes no43 7
yes 92 882
no
conf.mat.plot( predict_knn, actual_test )
To report Mean Squared Error (MSE):
mse( predict_knn, actual_test )
1] 0.09667969 [
The predictors that we used in the previous part, do not have the same scale. For example, variable day.mins
change between 0 and 351.5, whereas variable voice.plan
is binary. In this case, the values of variable day.mins
will overwhelm the contribution of voice.plan
. To avoid this situation we use normalization. So, we use min-max normalization and transfer the predictors as follows:
= kNN( formula, train = train_set, test = test_set, k = 8, transform = "minmax" ) predict_knn_trans
To report Confusion Matrix:
conf.mat.plot( predict_knn_trans, actual_test )
conf.mat.plot( predict_knn, actual_test )
To report the ROC curve, we need the probability of our classification prediction. We can have it by using:
= kNN( formula, train = train_set, test = test_set, k = 8, type = "prob" )[ , 1 ]
prob_knn
= kNN( formula, train = train_set, test = test_set, transform = "minmax", k = 8, type = "prob" )[ , 1 ] prob_knn_trans
To visualize the model performance between the raw data and the transformed data, we could report the ROC curve plot as well as AUC (Area Under the Curve) by using the plot.roc
function from the pROC package:
= roc( actual_test, prob_knn )
roc_knn = roc( actual_test, prob_knn_trans )
roc_knn_trans
ggroc( list( roc_knn, roc_knn_trans ), size = 0.8 ) +
theme_minimal() + ggtitle( "ROC plots with AUC") +
scale_color_manual( values = c( "red", "blue" ),
labels = c( paste( "AUC=", round( auc( roc_knn ), 3 ), "; Raw data; " ),
paste( "AUC=", round( auc( roc_knn_trans ), 3 ), "; Transformed data" ) ) ) +
theme( legend.title = element_blank() ) +
theme( legend.position = c( .7, .3 ), text = element_text( size = 17 ) ) +
geom_segment( aes( x = 1, xend = 0, y = 0, yend = 1 ), color = "grey", linetype = "dashed" )
To find out the optimal value of k
based on Error Rate, for the different values of k from 1 to 30, we run the k-nearest neighbor for the test set and compute the Error Rate for these models, by running kNN.plot()
command
kNN.plot( formula, train = train_set, test = test_set, transform = "minmax",
k.max = 30, set.seed = 3 )
The plot shows that the minimum value of Error Rate is for the case that k is 13; the smaller values of Error Rate indicates better predictions.