RM 5 Store/Retrieve Operators not Saving/Loading Properly
dragoljub
New Altair Community Member
Hi Everyone,
Using Rapid Miner 5.0 Beta:
I have a fairly large CSV file (2100 samples by 320 features) consisting of positive and negative integers with a range of -8 to +8 and with most entries set to 0. Read CSV properly reads in the file and the meta data view shows no missing entries. After saving the ExampleSet to a repository and loading it back I see many columns contain missing entries replaced by "?". It seems like Store does not properly save the data file and missing entries are created. It looks like values of "-1" seem to be getting dropped intermittently, while other negative numbers are saved fine.
I have loaded the CSV files using matlab and excell and everything works fine. I have also reproduced this problem by saving an ExampleSet file to CSV using RapidMiner than loading the CSV file into a repository file (.IOO) followed by a retrieve resulting in missing entries.
Hopefully this is just a little text to int glitch. A random CSV matrix of +/- integers may also reproduce this behavior.
-Gagi
Using Rapid Miner 5.0 Beta:
I have a fairly large CSV file (2100 samples by 320 features) consisting of positive and negative integers with a range of -8 to +8 and with most entries set to 0. Read CSV properly reads in the file and the meta data view shows no missing entries. After saving the ExampleSet to a repository and loading it back I see many columns contain missing entries replaced by "?". It seems like Store does not properly save the data file and missing entries are created. It looks like values of "-1" seem to be getting dropped intermittently, while other negative numbers are saved fine.
I have loaded the CSV files using matlab and excell and everything works fine. I have also reproduced this problem by saving an ExampleSet file to CSV using RapidMiner than loading the CSV file into a repository file (.IOO) followed by a retrieve resulting in missing entries.
Hopefully this is just a little text to int glitch. A random CSV matrix of +/- integers may also reproduce this behavior.
-Gagi
Tagged:
0
Answers
-
Hi,
thank you for reporting this small little, but very annoying bug. Integer was erroneously saved the same way as nominal attributes, where missings are mapped on -1. This will of course be solved in the final version.
Greetings,
Sebastian0 -
Thanks for the quick reply. Any word on when the next 5.0 version will be available? ;D
0 -
Hi,
the final release is dated in the mid December.
Greetings,
Sebastian0 -
Hi,
when I store an some data (7000 rows 15 colums with values between 0 an 1) with RM5.0.001 some values transform to 0 (even a complete column).
I tried different formats (Access, CSV, Excel), I even transformed the decimal seperator "," into "." but with no better result.
Using "Read Excel" worked fine although the meta data tip was not completely correct.
Greetings,
MaHo
0 -
Hi Maho,
You are still working with a non final version dated a few months ago. I would suggest installing the newest version from source forge. If you wait until tomorrow, it will be the newest update release, we are going to upload in a few hours. We have solved the bugs from the pre final versions until the final release and still are going to improve the software.
Greetings,
Sebastian
0
