721.txt

This complete matlab for neural network

发信人: GzLi (笑梨), 信区: DataMining
标  题: [合集]请问Apriori和FP算法到底那个效果好在large database
发信站: 南京大学小百合站 (Sat Aug 31 21:42:29 2002), 站内信件

highso (漫步者) 于Wed Aug 21 11:19:13 2002提到：

rt


jimo (寂寞) 于Wed Aug 21 13:00:30 2002提到：

呵呵
哪个好自己实现一下不就行了



mining (key) 于Wed Aug 21 21:13:28 2002)
提到：

Since the FP algorithm is developed for  large data sets, as to a very large d
ate collection,FP algorithm shoud work much faster than Apriori one


luoes (luoes) 于Fri Aug 23 20:12:30 2002提到：

自然是fp-tree更好，它只需遍历一次数据库，而apriori算法要N次
在大数据集上自然fp－tree好，



jimo (寂寞) 于Fri Aug 23 21:35:17 2002提到：

哎哎
但是fptree 还要递归呢



fervvac (高远) 于Sat Aug 24 11:25:23 2002提到：

Acoording to Prof. Han, their latest OP (opportunistic pruning) algorithms 
is the fatest one for tasks that mine the whole patterns. A recent work 
in ICDM02 could be a rival, as it beats FP by a large margin.



minerboy (miner) 于Tue Aug 27 14:09:21 2002)
提到：

很多递归是在内存里面做的 
比每次扫描硬盘数据库要强的多，建议你看介绍fp的那个paper,里面有比较


jimo (寂寞) 于Fri Aug 30 13:46:05 2002提到：

哎哎
paper我看了好多遍了
实现都实现了
他的比较不过是那么几个数据库而已
何况我也没说递归导致很满啊



fervvac (高远) 于Fri Aug 30 15:45:40 2002提到：

It is said that the recrusive constrution of the subtree will consume a lot
of memory though.

Ther are already a few improved algorithms after the introduction of fptree.
You guys, if keen on dm, should definitely look at the latest results, :)



jimo (寂寞) 于Fri Aug 30 16:31:27 2002提到：

很难有什么突破 了



fervvac (高远) 于Sat Aug 31 11:30:11 2002提到：

Well, one lesson I larned is that there is always niche if you are really
into the problem.

A shortcut to the identification of the hidden problems is to implement and
play with the state-of-the-art algorithms by yourself. In this sense, it is
a sad fact that many guys even don't want to implement Apriori by themself!