GSP - wrong results?

Anna
New Altair Community Member
The max-gap in the GSP operator does not seem to be working well.
I tried the GSP operator on a small data file:
IP | Timestamp | Page |
170 | 10 | a |
170 | 20 | c |
170 | 30 | e |
170 | 40 | f |
17 | 10 | a |
17 | 15 | c |
17 | 20 | f |
116 | 10 | a |
116 | 30 | c |
116 | 50 | e |
116 | 70 | d |
185 | 10 | a |
185 | 20 | c |
185 | 30 | f |
185 | 40 | e |
185 | 50 | b |
185 | 60 | e |
When I run it with window=0, min gap=1, max gap=1000, sup=0.5 I get the right results:
GSPSet
1.000: <Page = a> <Page = c>
0.750: <Page = a> <Page = e>
0.750: <Page = a> <Page = f>
0.750: <Page = c> <Page = e>
0.750: <Page = c> <Page = f>
0.750: <Page = a> <Page = c> <Page = e>
0.750: <Page = a> <Page = c> <Page = f>
when I run it with window =0, min gap=1, max gap=10, sup = 0.5, I get:
GSPSet
1.000: <Page = a> <Page = c>
0.750: <Page = a> <Page = e>
0.750: <Page = a> <Page = f>
0.750: <Page = c> <Page = e>
0.750: <Page = c> <Page = f>
and the right result should be:
GSPSet
0.750: <Page = a> <Page = c>
0.50: <Page = c> <Page = f>
For example pattern <a, e> is not supporetd at all because in case of 170 the gap is 20, in case of 116 the gap is 40, and in case of 185 the gap is 30 and 50. So each of the gaps is larger than max gap of 10.
Is this operator not implemented well in Rapid Miner?
Tagged:
0