Gaussian Naive Bayes Classifier

A simple C++ Gaussian Naive Bayes Classifier (GNBC) library using Eigen 3.3.7. Make sure to read the short-readme for its usage.

Dependecies

• Eigen 3.3.7

Installation

Include header file and compile .cpp file along with your "main.cpp" file.
You may need to change the include path of your Eigen library.

Usage

You can test the library by compiling test.cpp (with bayes.cpp).

Training

Declare a class object-classifier and train it using the following line:

GNBC classifierObject(trainDataX,trainDataY);

or

GNBC classifierObject();
classifierObject.train(trainDataX,trainDataY);

Classifying

• To classify a Eigen::VectorXf observation (vector that contains a single observation with length = numberOfFeatures), use:
  Eigen::VectorXf prediction = classifierObject.predict(observation);
int label = classifierObject.label(prediction);

  prediction: Vector that contains probabilities of each class (length = numberOfClasses)
  label: Contains index of class predicted (index of prediction vector that corresponds to the biggest probability)

• To classify a Eigen::MatrixXf observations (matrix that contains n observations with dimension = n*numberOfFeatures), use:
  Eigen::MatrixXf predictions = classifierObject.predictMatrix(observations);
Eigen::VectorXi labels = classifierObject.labelMatrix(predictions);

  predictions: MatrixXf that contains probabilities for every observation for each class (with dimension = n*numberOfClasses)
  labels: Contains indexes of classes predicted (indexes of predictions matrix that corresponds to the biggest probability row-wise)

You can view the test.cpp file for an example.

⚠️ At training stage, trainDataY (integer vector containing outcomes of examples given) should follow zero-based numbering!