"Broken SVM in 4.3"

noah977

Hello,

I've come across a strange problem. 

I'm attempting to train an SVM on a fairly simple data set of about 1800 records.

RM runs for about 3 seconds and returns a result with 0 weights and new vectors.  This must be some kind of error, but I can't seem to figure it out.

This intent is to use a Regression SVM to build a model based around the score. (column 2)

Can anyone help me figure out what I'm doing wrong?


Here is the XML:

<?xml version="1.0" encoding="MacRoman"?>
<process version="4.3">

  <operator name="Root" class="Process" expanded="yes">
      <operator name="CSVExampleSource" class="CSVExampleSource">
          <parameter key="filename"	value="/Users/noah/test_data.csv"/>
          <parameter key="id_column"	value="1"/>
          <parameter key="label_column"	value="2"/>
      </operator>
      <operator name="LibSVMLearner" class="LibSVMLearner">
          <parameter key="svm_type"	value="nu-SVR"/>
      </operator>
  </operator>

</process>

Here are a few rows of the actual data:

id, score, a, b, c, d, e, f, g, h, i, j, k, l, m, n, o, p, q, jockey_total, r, s, t, u
1, 0.975736568457561, 2, 11, , 0.987633891484376, 161, 499, 515, 454, 456, 544, 571, 133, 0.292868096301349, 0.333333333333333, 0.666666666666667, 3, 0.131984829329962, 3955, 0.0983037779491133, 2594, 0.666666666666667, 0
2, 0.964041095890434, 6, 1, , 0.969326167933506, 22, 499, 515, 454, 456, 544, 471, 126, 0.0146299104940125, 0, 0, 19, 0.0890924229808493, 3603, 0.0685431326557206, 3837, 0.315789473684211, 0.2
3, 0.964041095890434, 7, 5, , 0.958946059460525, 26, 499, 515, 454, 456, 544, 536, 120, 0.0146676586838751, 0.333333333333333, 0.333333333333333, 6, 0.0189054726368159, 1005, 0.072706065318818, 2572, 0.166666666666667, 0
4, 0.969018932874375, 5, 12, , 0.976029482968252, 56, 499, 515, 454, 456, 544, 814, 123, 0.0135066852021626, 0, 0.285714285714286, 7, 0.0570175438596491, 456, 0.0820323587851263, 3523, 0.142857142857143, 0
5, 0.974048442906594, 3, 4, , 0.970440990616629, 263, 499, 515, 454, 456, 544, 520, 122, 0.0139395866813886, 0, 0.0909090909090909, 11, 0.128205128205128, 858, 0.0970921636274027, 4058, 0.0909090909090909, 0
6, 0.970689655172434, 4, 6, , 0.976713816212313, 26, 499, 515, 454, 456, 544, 602, 125, 0.000679292864925718, 0.0666666666666667, 0.233333333333333, 90, 0.0526030368763558, 3688, 0.0712796998749479, 2399, 0.533333333333333, 6
7, 0.979130434782631, 1, 2, , 0.979130434782631, 0, 499, 515, 454, 456, 544, 471, 125, 0.000688918975681972, 1, 1, 1, 0.169156237383932, 4954, 0.0892917086270561, 2979, 1, 0
8, 0.978417266187054, 5, 4, , 0.976254744832316, 35, 499, 545, 454, 456, 544, 471, 114, 0.000701077430625211, 0.0625, 0.0625, 16, 0.0333333333333333, 1440, 0.0675074183976261, 1348, 0.0625, 6
9, 0.980769230769235, 2, 11, , 0.979336489366026, 217, 499, 545, 454, 456, 544, 727, 125, 0.000694564685698117, 0, 0.2, 5, 0.169156237383932, 4954, 0.0641821946169772, 1449, 0.2, 0
10, 0.979591836734694, 3, 9, , 0.980340323609057, 22, 499, 545, 454, 456, 544, 605, 128, 0.000655882933532074, 0.127659574468085, 0.319148936170213, 47, 0.0613586559532505, 1369, 0.0774693350548741, 3098, 0.340425531914894, 0.857142857142857

land

Hi,
the SVM does not support missing data. Your attribute g does not contain any information. If you use this process setting instead it will work:

<operator name="Root" class="Process" expanded="yes">
    <operator name="CSVExampleSource" class="CSVExampleSource">
        <parameter key="filename"	value="C:\Dokumente und Einstellungen\sland\Desktop\Noname1.txt"/>
        <parameter key="id_column"	value="1"/>
        <parameter key="label_column"	value="2"/>
    </operator>
    <operator name="AttributeFilter" class="AttributeFilter">
    </operator>
    <operator name="LibSVMLearner" class="LibSVMLearner">
        <parameter key="svm_type"	value="nu-SVR"/>
    </operator>
</operator>

Greetings,
  Sebastian