1
$\begingroup$

Hi there smart people,

I am new to data Science and wanted to take my first few steps. Unfortunately I struggle to find datasets or any data at all regarding my topics of interest.

For example, I wanted to build a simple program that takes a person's height and predicts the likely height of his/her child. (greetings from Galton)

But even for this relatively famous topic I am unable to find data.

Are there any good websited or something that list a lot of datasets for specific topics? A database of datasets, so to speak.

I mean, I am on the Internet, it should not be that difficult to find some data ;)

$\endgroup$
1
$\begingroup$

The Galton Height dataset seems to be exactly what you're looking for.

There are a ton of repositories of open datasets online. For beginners, I often see the UCI Machine Learning Repository. It has quite a few datasets which are easy to work with.

Some other resources for datasets:

  • Kaggle - hosts data science competitions, but you can download all the datasets without participating in a competition.
  • Data.gov - You can freely download data from US govt agencies
  • AWS Open Data - You need a AWS account, but these datasets are free-to-download
$\endgroup$
  • 1
    $\begingroup$ Hi there, thanks alot, a actually endet up on Kaggle which seems like a good fit for starting out. Iam suprised that the datasets arent bigger, especially the Galton Dataset is quiet small and old. Arent way bigger sets necessary for predictions?Off course more is always better but isnt there a minimum amount? Also my Take away is there isnt one big maine source everybody uses and I just missed. Its always kind of a lucky game to find the right data? $\endgroup$ – G.M Sep 6 at 6:09
  • $\begingroup$ @G.M The size of the dataset needed to train an effective model is determined by the size of the model and the difficulty of the problem. Training a deep neural network to recognize images might require a very large dataset. Training a linear regression to predict the height of offspring from their parents may require only a few dozen examples. $\endgroup$ – zachdj Sep 9 at 13:01
  • $\begingroup$ Also, you're correct that there isn't a single large repository of data used by everyone. In the research sphere, there are often benchmark datasets for certain classes of problems, and these datasets are usually easy to find. In industry, much of the data collected is proprietary, and companies are rarely incentivized to share their raw data with the public. $\endgroup$ – zachdj Sep 9 at 13:37

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.