0
$\begingroup$

I have a dataset that contains 44804 instances, each with the attributes: diagnoses, age and quantities (same diagnoses includes). I want to see which diagnosis is most seen for a given age.

For example, I want to be able to say that a diagnosis of "measles" is seen mostly at the age of "10".

What should I do? Should I try to classify or cluster? Or is there another tool in Weka that I could use?

$\endgroup$

1 Answer 1

1
$\begingroup$

From your description of the problem you are looking at doing data exploration, not anything too complicated or in depth. You can do this in other applications too, but in WEKA I would recommend a few possibilities:

  1. In the Preprocess tab you can quickly visualize different attributes vs. the class attribute. You can also get summaries for each attribute such as type (numeric/nominal), distinct values, unique values, etc.;

  2. In the Preprocess tab you can also apply filters on an attribute or instance basis, such as discretizing numerical variables or converting nominal attributes into binary attributes; and

  3. In the Visualize tab you can visualize any two attributes against one another and expand in to explore the structure that exists in your dataset.

$\endgroup$
3
  • $\begingroup$ Thanks for your reply. From what i searched, I need to add a filter like u said. I already tried visualizing but i need a table output that shows decreasing results of age-quantity density. Which filter gives me that result ? $\endgroup$ Sep 5, 2017 at 12:50
  • 1
    $\begingroup$ @MetinTurşu: Weka is not my first choice when it comes to producing tables and I am not sure if there is a best (or even good) way of doing it. I would try R, Octave/MATLAB or Excel if that is what you want to produce. $\endgroup$
    – user77876
    Sep 5, 2017 at 13:22
  • $\begingroup$ I know i ve already made this with R (dplyr library). But i need to do this with Weka :/ Any recommendations with Weka solution ? $\endgroup$ Sep 5, 2017 at 13:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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