Hi, all here,
I am having a more considertaion about Data Mining plug-in algorithms. When we say we are going to embed a uesr plug-in algorithm, so what is the context for that ? I mean in which case then we thing we need to embed a user plug-in algortihm? I know when we say we are going to embed a user costomermized plug-in algorithm, it means we want something more costomized. But what kind of customized features are generally concerned? Is it independant for different market sectors?
I dont think we can just try to embed a plug-in algorithm then compete it with avaialble algorithms to see which one is with better prediction accuracy?
Would please someone here give me some guidances about that?
Essentially, you would use a customized plug-in algorithm when one of the built-in algorithms doesn't perform the task you want. There are a wide variety of DM algorithms, and we have nine built in. For example, you may want hierarchical clustering, or link analysis, or a classification algorithm particularly suited towards fraud. Another example is an algorithm that allows you to specify additional penalties for specific incorrect predictions.
Considering your last question, yes, you can create a plug-in algorithm and compare it with the built in's to see which has better accuracy
|||Hi, Jamie, thanks a lot for your very helpful guidance.
With best regards,
No comments:
Post a Comment