When should we use a Bernoulli Naive Bayes classifier? If x is a random variable and is Bernoulli distributed, it means that my variables are binary variables, so they can assume only two values. For simplicity, let’s say these values are 0 and 1, and their probabilities are given by this equation: the probability of x is p if x=1, and the probability of x is q if x = 0 . So, if you know the probability of p , you can calculate the probability of q as q = 1 - p , where the probability lies between 0 and 1, as p + q = 1 . Here, we are considering only one input variable, right? My dataset is a single-dimensional dataset. So, if my dataset is single-dimensional, we can easily calculate the probability of x = 1 as p , and the probability of x = 0 as q . Then calculate for the non-mammals. So, out of 13, 1 is there where my birth is yes. Out of 13, 10 are there where can fly is no. Out of 13, 3 are there where living in water is yes, and out of 13, 4 are there where have legs is no. So, if you find this product, you get this probability: 0.00042. So, we multiply this probability with the individual probability of mammals class and the non-mammals class. So, mammals class into this value, you will get the probability of the query for the given class as mammals. Then, if you multiply this product, the probability of non-mammals, with this product that we calculated here, right, we get this probability, right? So, here is we are getting 0.0027. So, if you observe, student, if you observe, we check for the maximum probability. So, probability of attribute for the class level mammals into probability of mammals is greater than the probability of attributes for the non-mammals in the probability of non-mammals. So, because this probability is greater, my new query belongs to mammals. So, if you observe, the formula is the same as the we discussed in the previous sessions, right? Naive Bayes classifier only, my attributes are here the binary attributes. Now we can apply here the Bernoulli Naive Bayes classifier because my dataset is a binary dataset. So, what is the inbuilt function available in the scikit-learn library? Here, from `sklearn.naive_bayes` import `BernoulliNB`. So, here you know that in the Naive Bayes family, there are multiple variations, right? So, the first variation we are discussing here is Bernoulli. Right, so my Bernoulli Naive Bayes classifier is defined under this library `sklearn.naive_bayes`, right? So here, **B** is capital, and **NB** is capital. Once you have imported the required library, you can create the object of that class. So, before that, we are importing here the `train_test_split` function from the `model_selection` library to divide the dataset into training and testing sets. So, you know that this is the syntax for dividing the data into training and testing sets. Here, my test size is passed as 0.25, meaning 25 percent will be taken into the test dataset, and the remaining 75 percent into the training dataset. Then, we are creating here the object of the Bernoulli Naive Bayes classifier, right? Then `bnb` is the name of my object, right? With the help of this object, you can fit your model. You can execute your Bernoulli Naive Bayes classifier with the help of the `fit` function, where we are passing the `X_train` and `y_train` because my Naive Bayes classifier is a supervised learning technique. We need **X** and the corresponding **y** label for that **X** record. So, with the help of this training dataset, my model will build here a trend that we are using to predict the class of the new query. So, here is the test dataset in the **X_test** and **y_test** variables that we are passing to the `score` function, right? This `score` function we are calling with this object `bnb`, which will calculate the accuracy of this testing dataset. So, we are getting here 85 percent accuracy. Right? So, this `score` function is equivalent to the `accuracy_score` function. The inbuilt `accuracy_score` function is there under this `score` function.