void testGeneratedData20()
{
DataFrame df;
generatedData20(df);
GreedyStepwiseSearch uut;
shared_ptr<ConsistencySubsetEvaluator> evaluator(new ConsistencySubsetEvaluator());
uut.setEvaluator(evaluator);
// this has been validated against weka
vector<string> result = uut.findSubset(df);
CPPUNIT_ASSERT_EQUAL((int)result.size(), 3);
// order isn't important here, but easy
CPPUNIT_ASSERT_EQUAL(result[0], string("age"));
CPPUNIT_ASSERT_EQUAL(result[1], string("elevel"));
CPPUNIT_ASSERT_EQUAL(result[2], string("car"));
generatedData1000(df);
// these results have been validated against weka
result = uut.findSubset(df);
CPPUNIT_ASSERT_EQUAL((int)result.size(), 5);
// order isn't important here, but easy
CPPUNIT_ASSERT_EQUAL(result[0], string("salary"));
CPPUNIT_ASSERT_EQUAL(result[1], string("commission"));
CPPUNIT_ASSERT_EQUAL(result[2], string("elevel"));
CPPUNIT_ASSERT_EQUAL(result[3], string("car"));
CPPUNIT_ASSERT_EQUAL(result[4], string("zipcode"));
}
void testGeneratedData1000()
{
DataFrame df;
generatedData1000(df);
GreedyStepwiseSearch uut;
shared_ptr<CfsSubsetEvaluator> evaluator(new CfsSubsetEvaluator());
uut.setEvaluator(evaluator);
// this has been validated against weka
vector<string> result = uut.findSubset(df);
// for (unsigned int i = 0; i < result.size(); i++)
// {
// cout << result[i] << endl;
// }
CPPUNIT_ASSERT_EQUAL((int)result.size(), 1);
// order isn't important here, but easy
CPPUNIT_ASSERT_EQUAL(result[0], string("salary"));
}
void testGeneratedData20()
{
DataFrame df;
generatedData20(df);
GreedyStepwiseSearch uut;
shared_ptr<CfsSubsetEvaluator> evaluator(new CfsSubsetEvaluator());
uut.setEvaluator(evaluator);
// this has been validated against weka
vector<string> result = uut.findSubset(df);
CPPUNIT_ASSERT_EQUAL((int)result.size(), 1);
// This is slightly different from the Weka output. Weka lists both age and car, but also
// says the highest score is .619, the score for age alone, not the score for car and age.
// I wonder if they have some extra logic in there apart from what is in the paper. By
// removing car from the list Weka puts in the next highest correlated variable that is not
// zero. It may be a bug in my code, but I can't find it. It is reassuring that the 1000
// test is spot on.
CPPUNIT_ASSERT_EQUAL(result[0], string("age"));
}
