Big Data Analytics - Naive Bayes Classifier

Naive Bayes is a probabilistic technique for constructing classifiers. The characteristic assumption of the naive Bayes classifier is to consider that the value of a particular feature is independent of the value of any other feature, given the class variable.

Despite the oversimplified assumptions mentioned previously, naive Bayes classifiers have good results in complex real-world situations. An advantage of naive Bayes is that it only requires a small amount of training data to estimate the parameters necessary for classification and that the classifier can be trained incrementally.

Naive Bayes is a conditional probability model: given a problem instance to be classified, represented by a vector x = (x1, …, xn) representing some n features (independent variables), it assigns to this instance probabilities for each of K possible outcomes or classes.

$$p(C_k|x_1,....., x_n)$$

The problem with the above formulation is that if the number of features n is large or if a feature can take on a large number of values, then basing such a model on probability tables is infeasible. We therefore reformulate the model to make it simpler. Using Bayes theorem, the conditional probability can be decomposed as −

$$p(C_k|x) = \frac{p(C_k)p(x|C_k)}{p(x)}$$

This means that under the above independence assumptions, the conditional distribution over the class variable C is −

$$p(C_k|x_1,....., x_n)\: = \: \frac{1}{Z}p(C_k)\prod_{i = 1}^{n}p(x_i|C_k)$$

where the evidence Z = p(x) is a scaling factor dependent only on x1, …, xn, that is a constant if the values of the feature variables are known. One common rule is to pick the hypothesis that is most probable; this is known as the maximum a posteriori or MAP decision rule. The corresponding classifier, a Bayes classifier, is the function that assigns a class label $\hat{y} = C_k$ for some k as follows −

$$\hat{y} = argmax\: p(C_k)\prod_{i = 1}^{n}p(x_i|C_k)$$

Implementing the algorithm in R is a straightforward process. The following example demonstrates how train a Naive Bayes classifier and use it for prediction in a spam filtering problem.

The following script is available in the bda/part3/naive_bayes/naive_bayes.R file.

# Install these packages 
pkgs = c("klaR", "caret", "ElemStatLearn") 

# Split the data in training and testing 
inx = sample(nrow(spam), round(nrow(spam) * 0.9)) 
train = spam[inx,] 
test = spam[-inx,]  

# Define a matrix with features, X_train 
# And a vector with class labels, y_train 
X_train = train[,-58] 
y_train = train$spam  
X_test = test[,-58] 
y_test = test$spam  
# Train the model 
nb_model = train(X_train, y_train, method = 'nb',  
   trControl = trainControl(method = 'cv', number = 3)) 

# Compute  
preds = predict(nb_model$finalModel, X_test)$class 
tbl = table(y_test, yhat = preds) 
sum(diag(tbl)) / sum(tbl) 
# 0.7217391 

As we can see from the result, the accuracy of the Naive Bayes model is 72%. This means the model correctly classifies 72% of the instances.