Classification Example with Neuralnet in R

   The neural network models are widely used in regression, classification, and other types of analysis. It is a core principle of deep learning. Based on the human brain process, the neural network algorithm loosely imitates the learning method of biological neural networks.
   In this post, we'll briefly learn how to classify the Iris dataset with the 'neuralnet' package in R. The tutorial covers:

1. Preparing the data
2. Defining the model
3. Prediction and accuracy check
4. Source code listing

We'll start by loading the 'neuralnet' and 'caret' packages in R.

library(neuralnet)
library(caret)

Preparing the data

We'll load the Iris dataset and check the content of it.

data("iris")
str(iris)
'data.frame': 150 obs. of  5 variables:
 $ Sepal.Length: num  5.1 4.9 4.7 4.6 5 5.4 4.6 5 4.4 4.9 ...
 $ Sepal.Width : num  3.5 3 3.2 3.1 3.6 3.9 3.4 3.4 2.9 3.1 ...
 $ Petal.Length: num  1.4 1.4 1.3 1.5 1.4 1.7 1.4 1.5 1.4 1.5 ...
 $ Petal.Width : num  0.2 0.2 0.2 0.2 0.2 0.4 0.3 0.2 0.2 0.1 ...
 $ Species     : Factor w/ 3 levels "setosa","versicolor",..: 1 1 1 1 1 1 1 1 1 1 ...

Next, we'll split the dataset into the train and test parts.

set.seed(123)

indexes=createDataPartition(iris$Species, p=.85, list = F)
train = iris[indexes, ]
test = iris[-indexes, ] 

We'll extract x and y label parts from the test data to compare the predicted results later. Here column 5 is Species that is a label data.

xtest = test[, -5]
ytest = test[, 5]

Defining the model

We'll define the model with the 'neuralnet' function and fit it on train data. We'll train the function with the below parameters.

nnet=neuralnet(Species~., train, hidden = c(4,3), linear.output = FALSE)

The hidden parameter in the neuralnet function specifies the number of hidden neurons in each layer. You can change the inputs. However, it takes time to train the model with a large number of neurons.
We can plot the fitted model and check what is inside the model.

plot(nnet)

Prediction and accuracy check

Next, we'll predict the test data with the model.

ypred = neuralnet::compute(nnet, xtest)
 

yhat = ypred$net.result

print(yhat)

          [,1]         [,2]         [,3]
3   1.000000e+00 1.345375e-14 1.346945e-64
17  1.000000e+00 1.342452e-14 1.349464e-64
25  1.000000e+00 1.345269e-14 1.347036e-64
32  1.000000e+00 1.342934e-14 1.349048e-64
36  1.000000e+00 1.345266e-14 1.347038e-64
69  3.233607e-23 1.000000e+00 9.862038e-11
70  2.575823e-17 1.000000e+00 1.687434e-20
71  2.171171e-35 4.022875e-05 9.998378e-01
82  2.499568e-17 1.000000e+00 1.040542e-20
96  2.635722e-17 1.000000e+00 4.848410e-20
101 1.622106e-55 8.186289e-23 1.000000e+00
107 5.660145e-48 5.500966e-14 1.000000e+00
122 1.924033e-53 6.920406e-20 1.000000e+00
125 1.017510e-54 2.793636e-21 1.000000e+00
129 2.989258e-55 5.776680e-22 1.000000e+00 

The prediction result shows the probability of each class. We need the extract the class with the highest prediction values as the predicted result.

yhat=data.frame("yhat"=ifelse(max.col(yhat[ ,1:3])==1, "setosa",
                       ifelse(max.col(yhat[ ,1:3])==2, "versicolor", "virginica")))

Finally, we'll check the prediction accuracy with the confusion matrix function.

cm = confusionMatrix(as.factor(ytest), yhat$yhat)

print(cm)
Confusion Matrix and Statistics

            Reference
Prediction   setosa versicolor virginica
  setosa          5          0         0
  versicolor      0          4         1
  virginica       0          0         5

Overall Statistics
                                          
               Accuracy : 0.9333          
                 95% CI : (0.6805, 0.9983)
    No Information Rate : 0.4             
    P-Value [Acc > NIR] : 2.523e-05       
                                          
                  Kappa : 0.9             
 Mcnemar's Test P-Value : NA              

Statistics by Class:

                     Class: setosa Class: versicolor Class: virginica
Sensitivity                 1.0000            1.0000           0.8333
Specificity                 1.0000            0.9091           1.0000
Pos Pred Value              1.0000            0.8000           1.0000
Neg Pred Value              1.0000            1.0000           0.9000
Prevalence                  0.3333            0.2667           0.4000
Detection Rate              0.3333            0.2667           0.3333
Detection Prevalence        0.3333            0.3333           0.3333
Balanced Accuracy           1.0000            0.9545           0.9167

   In this tutorial, we've briefly learned how to classify data with 'neuralnet' in R. The full source code is listed below.


Source code listing

library(neuralnet)
library(caret) 

data("iris")
str(iris) 

set.seed(123)

indexes=createDataPartition(iris$Species, p=.85, list = F)
train = iris[indexes, ]
test = iris[-indexes, ] 

xtest = test[, -5]
ytest = test[, 5] 

nnet=neuralnet(Species~., train, hidden = c(4,3), linear.output = FALSE)

plot(nnet) 

ypred = neuralnet::compute(nnet, xtest) 

yhat = ypred$net.result

print(yhat)

yhat=data.frame("yhat"=ifelse(max.col(yhat[ ,1:3])==1, "setosa",
                       ifelse(max.col(yhat[ ,1:3])==2, "versicolor", "virginica")))

cm=confusionMatrix(as.factor(ytest), yhat$yhat)
print(cm)