Features

New KDnuggets Poll is asking:

What will usually give better improvement in data mining results:

• Adding more data
• Using more advanced algorithms
• It depends (please comment)

Of course a lot depends on the specifics, but I think it is a useful rule of thumb:

When building predictive models, adding more (relevant) data usually gives more of an improvement, especially in the initial phases of the project, than using a more advanced algorithm.

Here is a recent example: More data usually beats better algorithms, by Anand Rajaraman.