ds/TestKNN.cpp

This is an example on how to use the KNN class for classification.

#include "../Test.h"
#include <math/ds/DataSet.h>
#include <math/ds/KNN.h>
#include <math/ds/utils.h>
 
 
class KNNTestCase : public Test
{
  bool TestConstructor() {
    auto clf = KNN(1);
    return true;
  }
 
  bool TestFit() {
    Matrix<double> input_data(
    { { 2, 1 }, { 4, 1 }, { 2, 3 }, { 4, 3 }, { 8, 6 }, { 10, 8 }, { 2, 12 }, { 2, 14 }, { 4, 14 } });
    Matrix<double> labels({ { 0 }, { 0 }, { 0 }, { 0 }, { 1 }, { 1 }, { 2 }, { 2 }, { 2 } });
 
    auto clf = KNN(2);
    clf.fit(input_data, labels);
 
    AssertEqual(clf.weights, input_data);
 
    clf = KNN(2);
    clf.fit(input_data, labels);
 
    AssertEqual(clf.weights, input_data);
 
    return true;
  }
 
  bool TestPredict() {
    Matrix<double> input_data(
    { { 2, 1 }, { 4, 1 }, { 2, 3 }, { 4, 3 }, { 8, 6 }, { 10, 8 }, { 2, 12 }, { 2, 14 }, { 4, 14 } });
    Matrix<double> labels({ { 0 }, { 0 }, { 0 }, { 0 }, { 1 }, { 1 }, { 2 }, { 2 }, { 2 } });
    Matrix<double> test_data({ { 1, 0 }, { 8, 8 }, { 0, 1 }, { 3, 16 } });
    Matrix<double> expected_labels({ { 0 }, { 1 }, { 0 }, { 2 } });
 
    auto clf = KNN(2);
    clf.fit(input_data, labels);
 
    auto predictions = clf.predict(test_data);
    AssertEqual(predictions, expected_labels);
 
    return true;
  }
 
  bool TestPredictionOnIrisData() {
    Set dataset  = Set("../../tests/ds/train.tsv", 4, 1);
    Set test_set = Set("../../tests/ds/test.tsv", 4, 1);
 
    auto clf = KNN(10);
    clf.fit(dataset.Input, dataset.Output);
 
    auto preds = clf.predict(test_set.Input);
 
    auto acc = accuracy(preds, test_set.Output);
 
    std::cout << "Iris flower classification, achieved " << acc * 100. << "% accuracy." << std::endl;
    AssertEqual(acc, 1.0);
    return true;
  }
 
 
public:
  virtual void run() {
    TestConstructor();
    TestFit();
    TestPredict();
    TestPredictionOnIrisData();
  }
};
 
int main() {
  KNNTestCase().run();
  return 0;
}