發信人: yaomc (白頭翁&山東大漢), 信區: DataMining
標  題: [合集]非平凡過程？
發信站: 南京大學小百合站 (Thu Jan 17 19:32:16 2002), 站內信件

armen (理性瘋狂) 于Fri Dec 28 21:23:08 2001提到：

數據挖掘是從巨量數據中產生有效的、新穎的、潛在有用的、最終可理解的
模式的非平凡過程。
      ~~~~~~~~~~~~~什么叫做非平凡過程啊？什么又是平凡過程？
這詞怎么來的？


joe (孔明/3) 于Fri Dec 28 21:42:46 2001提到：

就是說你要挖出來，而不是撿出來。


armen (理性瘋狂) 于Fri Dec 28 21:55:37 2001提到：

懂你的意思了，有更詳細的解釋么？我多了解一點非平凡這個概念



roamingo (漫步鷗) 于Fri Dec 28 22:26:30 2001提到：

"A DM algothim has many loops and very complicated data structure."  
Is this another interpretation for non-trivial?



daniel (飛翔鳥) 于Fri Dec 28 22:40:55 2001提到：


non-trivial, i.e. not easy



yaomc (白頭翁&山東大漢) 于Sat Dec 29 09:48:36 2001提到：

I think the nontrivial perhaps means that process of DM does not equal the 
simply query through the database to seek knowledge, it need more steps 
and more skills.


helloboy (hello) 于Sat Dec 29 10:37:09 2001提到：

I think non-trival means datamining needs iterative process.
Preprocess,mining and evaluate patterns are iterative.
When we found out some patterns,perhaps the patterns are not
so good.So we need to preprocess again,mine again and evaluate
again.Until the result can satisfy us.



armen (理性瘋狂) 于Sat Dec 29 20:55:11 2001提到：

i think your interpretation is the same to joe's

i want to know what  the non-trivial process means in maths

does it just mean a not-easy process?



explorer (void) 于Sun Dec 30 10:06:49 2001提到：

我的理解是數決挖掘過程不是線性的，不是從開始一直向下走到結束。
在挖掘過程中有反復，有循環，有跳轉，而且這種反復和循環和跳轉是沒有規律的。
僅供參考。


yaomc (白頭翁&山東大漢) 于Sun Dec 30 11:19:57 2001提到：

平凡的東西是很容易得到的,也比較淺顯,或者說是可以比較準確的預測的。
獲得此類的知識不需要太多的技巧和應用專門的工具，只要對于此領域比較熟悉，
能夠熟練的預測事物的發展趨勢。
而非平凡則是相對于平凡來說的。數據挖掘有時候強調的是，所挖掘的知識
往往不易通過簡單的分析就能夠得到，這些知識可能隱含在表面現象的內里，
需要經常大量數據的比較分析，應用一些專門對付大數據量的工具，才有可能得到。
得到的知識往往具有出乎意料的意味，因此也往往是不容易預測到的，當然，
數據挖掘得到的知識也用于對事物趨勢的預測。
有時候數據挖掘的目的是發現那些出現概率比較小的現象，這些東西好像用一般
統計的方法往往很難獲得。

所以，俺認為數據挖掘得到的非平凡知識就是那些往往出乎預料的東西。不是那些領導


fervvac (高遠) 于Sun Dec 30 15:51:10 2001提到：

My understanding for this terms is somewhat similar to yours:

trivial means sth. that can be easily known, obtained, etc. 

Before DM came into play, peoples had already begun some analysis of the hist
oric data, but mostly by using some naive methods (like counting, drawing cur
ves, etc) or some basic statistical methods (finding the distribution, cross 
validatation). I guess 
those methods are called trivial in the DM context.

So what DM methods are trying to do is a step further. For example, for the s
tock data, previously we can only draw the curves, try to predict what's the 
trend solely by the experience of the analyst, but with DM techniques, we mig
ht do it more 
accurately and probably more scientifically, :-)

比較淺顯,或者說是可以比較準確的預測的。