Can anyone direct me to a good source?
I am after a data set with: 1) A numeric or Binary DV 2) At least 6 possible predictor variables 3) At least one categorical predictor variable 4) At least 100 cases
Any assistance would be greatly appreciated
-
Predictive modeling of what? (I have no idea what 'DV' means in this context ... I assume it's not Digital Video) – Joe Sep 16 '13 at 10:45
-
I guess DV stands for dependent variable. R package AER has plenty of datasets available that satisfy some or all the conditions. – johnny Sep 16 '13 at 11:03
-
1It would be helpful to understand what you want this to be predictive of (climate modeling, population models, biological sampling, ...) and some more details so the community can help you find what you need. – Jeanne Holm Sep 17 '13 at 10:46
If you're already using a Python toolchain, then the easiest option is probably skdata. See the data sets it provides.
Kaggle hosts predictive modeling competitions where you might find the data set you are looking for to test your model.
UCI hosts a Machine Learning Repository with many datasets that are labeled by about method, type, etc.
In R there is a package 'readMLData' to read these datasets directly (PDF details).
With Python you can use the sklearn package to load many of the datasets (link).
Have a look at the homepages of machine learning toolkits like WEKA. It has a subpage Datasets containing several collections of datasets. I went for the first one, datasets-UCI.jar, renamed to .zip to extract the individual datasets and opened the first one, anneal.arff, in a text editor:
- Nominal or numeric class (not sure from description)
- 38 possible predictor variables: 3 integer, 29 nominal, 6 continuous
- Check.
- 798 instances.
