(English) How to do Predictive Analytics – Part 3

对不起，此内容只适用于English。

评论

  1. Just found this blog, and am favorably impressed! I especially like the amplification of the CRISP-DM process related to data. I’ve found the exact same problem with the data at some companies: they think they have data, but when you look, either (1) there really isn’t any data populated that is interesting, or (2) because the warehouse wasn’t built with data mining in mind, you can’t join together the tables (and hence the key fields) you would like for data mining.

    As with most of the data mining process, it is iterative and you must continually look ahead and then back. Of course, this makes explaining the process to novices more challenging.

    Dean Abbott

  2. Thanks for the amplification Dean. I certainly concur with your second point as well. I can think of quite a few cases where we have had to go back to the operational data feeds coming into the warehouse to get at some of the data for mining. In fact i’m on a project right now which is somewhat like that.

    John McConnell