Paul Chang created MADLIB-1111:
----------------------------------

             Summary: MADlib Random Forest EOFException when calling from JDBC
                 Key: MADLIB-1111
                 URL: https://issues.apache.org/jira/browse/MADLIB-1111
             Project: Apache MADlib
          Issue Type: Bug
            Reporter: Paul Chang


We are calling forest_train from Java via JDBC and we are receiving the 
following error:

[08006] An I/O error occurred while sending to the backend.
java.io.EOFException

We see this only with MADlib 1.9.1 and 1.10.0, but not 1.9.0.

Here's a script that reproduces the problem:

DROP TABLE IF EXISTS credit_test1;
CREATE TABLE credit_test1 (
  id INTEGER NOT NULL,
  times90dayslate INTEGER,
  revolving_util NUMERIC(19, 10),
  debt_ratio NUMERIC(19, 10),
  credit_lines BIGINT,
  monthly_income DOUBLE PRECISION,
  times30dayslate_2years INTEGER,
  srsdlqncy INTEGER
);
INSERT INTO credit_test1 (id, times90dayslate, revolving_util, debt_ratio, 
credit_lines, monthly_income, times30dayslate_2years, srsdlqncy)
VALUES
  (1, 0, 0.26, 0.324, 4, 1956.59, 0, 0),
  (1, 0, 0.46, 0.372, 7, 3595.37, 0, 0),
  (1, 0, 0.14, 0.690, 7, 2507.17, 0, 0),
  (1, 0, 0.32, 0.188, 7, 5158.11, 0, 0),
  (1, 0, 0.17, 0.343, 4, 1108.77, 0, 0),
  (1, 0, 0.37, 0.385, 5, 3893.31, 0, 0),
  (1, 0, 0.34, 0.452, 3, 2848.79, 0, 0),
  (1, 1, 0.47, 0.509, 7, 2158.23, 1, 1),
  (1, 0, 0.35, 0.104, 6, 4285.52, 0, 0),
  (1, 0, 0.30, 0.043, 4, 1334.18, 0, 0);
DROP TABLE IF EXISTS credit_aftest, credit_aftest_group, credit_aftest_summary;
SELECT "madlib"."forest_train"(
  /* training table */   'credit_test1',
  /* output table */     'credit_aftest',
  /* id col */           'id',
  /* dependent col */    'credit_lines',
  /* feature cols */     'revolving_util, debt_ratio, monthly_income',
  /* exclude cols */     CAST(NULL AS text),
  /* grouping cols */    CAST(NULL AS text),
  /* num trees */        10,
  /* num rand feat */    CAST(NULL AS integer),
  /* var importance */   FALSE,
  /* num permutations */ 1,
  /* max depth */        10,
  /* min split */        20,
  /* min bucket */       CAST(NULL AS integer),
  /* num splits */       10,
  /* surrogate param */  CAST(NULL AS text),
  /* verbose? */         FALSE,
  /* sample ratio */     1.0);




--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

Reply via email to