hello. in this page :
http://www.philippe-fournier-viger.com/spmf/index.php?link=datasets.php
Written "BMSWebView1 (Gazelle) ( KDD CUP 2000)" is for sequence pattern mining! i test it in spmf wit GSP, SPADE, CLOSPAN and etc, but do not have any output!!! why? please help me.
i want three Large DB for run:
GSP, SPADE, Prifixspan, clospan, BIDE, seq-dim
what databases i select that help me? motion it, i have at least two large DB.



Edited 2 time(s). Last edit at 04/14/2014 03:06PM by webmasterphilfv.

It depends on the parameter "minsup" that you are using.

For BMSWebView1, to get results, you need to set minsup very very low.

For example, for minsup = 0.01 you will get 77 sequential patterns with GSP

The lower you set minsup, the more patterns you will find.


For your experiments, you may use BMSWebView1, BIBLE, Kosarak, SIGN, LEVIATHAN, FIFA....

They are all suitable for sequential pattern mining.



Edited 2 time(s). Last edit at 04/14/2014 03:08PM by webmasterphilfv.

thank you. i want two DB that work with many min_sups!! (low min_sup and medum and larg) the DB that is good and complete. what is better?

The more you set minsup low, the more patterns you will get. And the number of patterns can increase exponentially when minsup is set lower. And when the number of patterns increase the algorithms get slower and consume more memory.

So if you set minsup too low, the algorithm will never terminate or run out of memory. And if you set minsup too high you will get no patterns or not enough.

Now, what is "too high" or "too low" depends on the dataset and the algorithm.

This picture from my PAKDD 2014 paper shows a comparison of the performance of the algorithm for the minsup values on the main datasets on the webpage:





Edited 1 time(s). Last edit at 04/15/2014 03:59AM by webmasterphilfv.

for MD-sequence mining, what am i do? for example for seq-dim? what DB i use?

For SeqDIM, there is no public dataset.