# What Is Naïve Bayes?

Naïve Bayes is an object classification algorithm. In this case, naïve means strong — as in, this classification algorithm supposes data point attributes are strongly independent. Some common use cases of naïve Bayes algorithms include:

• Medical diagnoses
• Text analyses
• Email spam filters

Naïve Bayes classification algorithms are widely adopted in machine learning contexts because implementation is easy.

Apply to a data science bootcamp with The Data Incubator today.

## Naïve Bayes Meaning

Naïve Bayes classification (also referred to as simply “Bayes” or “independent Bayes”) is so named because of the theorem with which it shares the name. Bayes’ Theorem states that event probability is adjustable as new information is fed into a system.

This type of Bayes classification is considered naïve because it assumes that data point attributes are autonomous and unrelated to one another.

For example, a naïve Bayes classification algorithm sorting images of various fruits would “understand” that an apple is red, circular, with a particular circumference range but wouldn’t make the distinction of each characteristic simultaneously. Strawberries are red also, oranges are also circular, etc.

The really neat thing about a naïve Bayes classification algorithm is that it isn’t just one algorithm by itself. Instead, it’s a set of algorithms running independently according to statistics. The algorithms are pretty simple to create and offer more efficient operation than some of the more complex Bayesian algorithms.

One of the most popular and common uses of naïve Bayesian algorithms is the simple spam filter you either love or love to hate. Your email provider programs the filter with specific keywords, and all incoming messages — unless the address is specifically saved in your address book — are automatically placed in the spam filter.

So, yes: naïve because the algorithm looks at attributes individually, but strong in its ability to learn and adapt as more information flows into the machine (such as when you move an email from your spam folder to your inbox or vice versa).

## Benefits of Naïve Bayes

The advantages of this classification algorithm include the following:

• It’s simple to create
• Implementation is rather straightforward
• It can handle continuously streaming data as well as isolated data
• It can easily scale with each additional predictor or data point
• It’s fast and agile and can make predictions in real-time
• It doesn’t scan for features or data that are not relevant to the classification system

## Drawbacks of Naïve Bayes

That’s not to say that naïve Bayes is perfect for all applications. It does have some disadvantages, such as:

• If you test the classification algorithm using variables not used when training the system, naïve Bayes assigns those variables a probability of zero, thus making no predictions regarding that specific data point.
• As good as it can be, it rather stinks at making estimations and assumptions.
• In its lousy estimation, it makes an…(*cough*)…it assumes that each feature it’s checking for is independent of all other features, which is, in fact, why its name is so apropos — in data science as in life, you’ll likely never run into a data set with features not dependent on one another.

## What are you waiting for?

Want to take a deep dive into the data science skills you need to become a successful data scientist? The Data Incubator has got you covered with our immersive data science bootcamp.