I'm Bayes, Naive Bayes (NB). I'm a successful classifier based upon the principle of maximum a posteriori (MAP). As a classifier, I am highly scalable, requiring a number of parameters linear in the number of variables (features/predictors) in a learning problem. I have several properties, for example, I am computationally faster, if you can hire me to classify something I'm simple to implement, and I can work well with high-dimensional datasets. Moreover, I can handle missing values in your dataset. Nevertheless, I'm adaptable since the model can be modified with new training data without rebuilding the model.
Sounds something similar to James Bond movies? Well, you/we can think a classifer as agent 007, right? Just kidding. I believe I am not as the parameters of the Naive Bayes classifier such as priori and conditional probabilities are learned or rather determined using a deterministic set of steps: this involves two very trivial operations that can be blindingly fast on modern computers, that is, counting and dividing. There is no iteration. There is no epoch. There is no optimization of a cost equation (which can be complex, of cubic order on an average or at least of square order complexity). There is no error back-propagation. There is no operation(s) involving solving a matrix equation. These make Naive Bayes and its overall training faster.
However, before hiring this agent, you/we can discover his pros and cons so that we can use this agent like a trump card by utilizing it's best only. Well, here's table summarizing the pros and cons of this agent:
| Agent | Pros | Cons | Better at |
|
Naive Bayes (NB) |
- Computationally fast - Simple to implement - Works well with high dimensions - Can handle missing values - Requires a small amount of data to train the model - Is adaptable since the model can be modified with new training data without rebuilding the model |
- Relies on independence assumptions and so performs badly if the assumption does not meet - Relatively low accuracy - If you have no occurrences of a class label and a certain attribute value together then the frequency-based probability estimate will be zero |
- When data has lots of missing values - When dependencies of features from each other are similar between features - Spam filtering and classification - Classifying a news article about technology, politics, sports, and so on. - Text mining |