[MADLIB-1219] RF: null_as_category=TRUE issues

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: v1.14
• Component/s: Module: Random Forest
• Assignee: Rahul Iyer
• Reporter: Frank McQuillan
• Created: 22/Mar/18 17:46
• Updated: 13/Apr/18 16:38
• Resolved: 26/Mar/18 18:25

Description

(1)
I cannot get null_as_category=TRUE to work when variable importance is used:

DROP TABLE IF EXISTS null_handling_example;
CREATE TABLE null_handling_example (
    id integer,
    country text,
    city text,
    weather text,
    response text
);
INSERT INTO null_handling_example VALUES
(1,null,null,null,'a'),
(2,'US',null,null,'b'),
(3,'US','NY',null,'c'),
(4,'US','NY','rainy','d');

RF:

DROP TABLE IF EXISTS train_output, train_output_group, train_output_summary;
SELECT madlib.forest_train('null_handling_example',  -- source table
                           'train_output',    -- output model table
                           'id',              -- id column
                           'response',        -- response
                           'country, weather, city',   -- features
                           NULL,              -- exclude columns
                           NULL,              -- grouping columns
                           2::integer,        -- number of trees
                           2::integer,        -- number of random features
                           TRUE::boolean,     -- variable importance
                           1::integer,        -- num_permutations
                           3::integer,        -- max depth
                           2::integer,        -- min split
                           2::integer,        -- min bucket
                           2::integer,        -- number of splits per continuous variable
                           'null_as_category=TRUE'
                           );

produces this error

ERROR:  plpy.SPIError: invalid array length
DETAIL:  array_of_float: Size should be in [1, 1e7], 0 given
CONTEXT:  Traceback (most recent call last):
  PL/Python function "forest_train", line 42, in <module>
    sample_ratio
  PL/Python function "forest_train", line 609, in forest_train
  PL/Python function "forest_train", line 1058, in _calculate_oob_prediction
PL/Python function "forest_train"

When variable importance is FALSE, it does not produce this error.

(2)
is null_as_category working for RF?

If I do get a tree trained, prediction seems wrong:

DROP TABLE IF EXISTS table_test;
CREATE TABLE table_test (
    id integer,
    country text,
    city text,
    weather text,
    expected_response text
);
INSERT INTO table_test VALUES
(1,'IN','MUM','cloudy','a'),
(2,'US','HOU','humid','b'),
(3,'US','NY','sunny','c'),
(4,'US','NY','rainy','d');

DROP TABLE IF EXISTS prediction_results;
SELECT madlib.forest_predict('train_output',
                             'table_test',
                             'prediction_results',
                             'response');
SELECT s.id, expected_response, estimated_response
FROM prediction_results p, table_test s
WHERE s.id = p.id ORDER BY id;

produces

 id | expected_response | estimated_response 
----+-------------------+--------------------
  1 | a                 | a
  2 | b                 | a
  3 | c                 | a
  4 | d                 | d
(4 rows)

but the same example for decision tree predicts properly.

Issue Links

• links to: GitHub Pull Request #249