1
$\begingroup$

is the survival table classification method on the Kaggle Titanic dataset an example of an implementation of Naive Bayes ? I am asking because I am reading up on Naive Bayes and the basic idea is as follows: "Find out the probability of the previously unseen instance belonging to each class, then simply pick the most probable class"

The survival table (http://www.markhneedham.com/blog/tag/kaggle/) seems like an evaluation of the possibilities of survival given possible combinations of values of the chosen features and I'm wondering if it could be an example of Naive Bayes in another name. Can someone shed light on this ?

$\endgroup$
2
$\begingroup$

Naive Bayes is just one of the several approaches that you may apply in order to solve the Titanic's problem. The aim of the Kaggle's Titanic problem is to build a classification system that is able to predict one outcome (whether one person survived or not) given some input data. The survival table is a training dataset, that is, a table containing a set of examples to train your system with.

As I mentioned before, you could apply Naive Bayes to build your classification system to solve the Titanic problem. Naive Bayes is one of the simplest classification algorithms out there. It assumes that the data in your dataset has a very specific structure. Sometimes Naive Bayes can provide you with results that are good enough. Even if that is not the case, Naive Bayes may be useful as a first step; the information you obtain by analyzing Naive Bayes' results, and by further data analysis, will help you to choose which classification algorithm you could try next. Other examples of classification methods are k-nearest neighbours, neural networks, and logistic regression, but this is just a short list.

If you are new to Machine Learning, I recommend you to take a look to this course from Stanford: https://www.coursera.org/course/ml

$\endgroup$
2
  • $\begingroup$ Thanks for the insight. You're right the survival table is just the training data set. I guess what confused me a little is the fact that the example just used the survival table to predict against the test data, without explicitly stating the algorithm used - it basically used the survival table as a lookup table to predict the outcome for each test case passenger. My question should have been whether this specific approach was an example of using Naive Bayes. $\endgroup$ – femibyte Feb 13 '15 at 9:19
  • $\begingroup$ No, it is not. It is even simpler than Naive Bayes; it looks like a simplified version of k-nearest neighbours classification. They group together or cluster the training examples. Each cluster contains the examples in the training set which have the same values in the input variables. Each cluster's class is computed as the rounded mean of the surviving value for the individuals in the cluster, and the results are stored in a lookup table. The lookup table assigns an output value to each test example depending on its input values. $\endgroup$ – Pablo Suau Feb 13 '15 at 10:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.