Unlike the other classification algorithms described in the other blog posts in this series, the naïve bayes classifier uses a straightforward statistical approach. It builds upon conditional probability theory, where an observation is classified by computing the likelihood of it belonging to a certain group.
Let’s say we want to know what type of platform a certain customer is most willing to use to find information about your company. They may use their computer to look online, use the browser on their smartphone to read the mobile site, or load your company’s own app.
Based on the attributes of the observed customer, for example a female between 20-30 years old with a college education, the naïve bayes approach will compute the probability of that customer using a phone application, computer website, or a mobile website and classify her into the class with the highest probability.
Naïve bayes is one of the tools used to do this, and it can be a helpful addition to our toolbox.
Although the algorithm is called naïve, it is a faster way to classify information about customers based on probability, rather than relying on huge data sets. This helps reduce to load on the AI system and keeps the output manageable and effective.
Since naïve bayes classifiers do not need a model to be trained, they are easy and quick to implement. This allows for testing on real data without spending a lot of time and money on developing the model. When implemented, they are quick to make predictions.
The naïve bayes algorithm does not need much data to perform well and can be used on both binary and multiple class predictions, which further reduces the time and money spent on modeling. However, as the name indicates, the classifier is naïve. It requires assumptions that do not always hold up, which may cause the system to perform poorly.
For our final classifier, we’ll look at neural networks. These are complex because data moves through them in elaborate ways, rather than the relatively simple systems we’ve seen so far.
If you want to read all the related articles on the topic of AI algorithms, here is the list of all blog posts in this article series: