Everything you need to know about Naive Bayesian Classifier!

Naïve Bayesian Classifier (NB Classifier) is an advanced classification technique that is a self-learning or machine learning technique that uses Bayes Theorem of conditional probability to predict the class of an observation. NB Classifier uses Bayes theorem of conditional probability assuming the predictor variables are independent. It also assumes the continuous variables to follow a normal distribution.

Application and Interpretation:

Naïve Bayes Classifier is a simple algorithm and works very well if the predictor variables are independent to each other. In the absence of large data, the algorithm some time work more efficiently than other complex classification algorithms.

This technique is more useful when it is more important to know the most likely class rather than knowing the actual probabilities for various classes. One example of successful application of this algorithm is in the applications of classifying a mail as spam or not.

Practice Dataset: https://learn.analyttica.com/simulations/Customer_Attrition_Naive_Bayes

Input:

In AnalytticaTreasureHunt, to run Naïve Bayes Classifier, you need to select the columns of the data you want to run the classifier on and then use the path: Machine Learning -> Classification — Others > Naïve Bayes Classifier.

You can select multiple variables, numeric or categorical, as predictor variables. The dependent or target variable is expected to be categorical. It is numeric, it assumes that the values are distinct categories. The target variable must have very limited unique values (preferably maximum 4). Neither ATH nor R will give an error otherwise. But the usefulness of the analysis will be lost or R engine may hang.

Output:

The output of the function will give the contingency table with respect to each predictor variable (if categorical) or mean & standard deviation of predictor variables. ATH will also give you the classification error, which will indicate the accuracy of the model.