Evaluation 🎯

This is the list of metrics used by StraighforwardNeuralNetwork to evaluate the neural network from Data.

• Accuracy
• Weighted Accuracy
• F1 score
• Mean Absolute Error
• Root-Mean-Square Error

Accuracy

Mainly used for classification and multiple classification.
The accuracy corresponds to the number of well-classified examples in the testing set of dataset. It is equivalent to precision.

Here the function to retreive the accuracy:

void StraightforwardNeuralNetwork::getGlobalClusteringRate() const;

See more detail on precision

Weighted Accuracy

Mainly used for classification.
The weighted accuracy corresponds to the number of well-classified examples in the testing set of dataset. It is equivalent to recall.

Here the function to retreive the weighted accuracy:

void StraightforwardNeuralNetwork::getWeightedClusteringRate() const;

See more detail on recall

F1 score

Mainly used for binary classification.
The F1 score is calculated from the precision and recall.

Here the function to retreive the weighted F1 score:

void StraightforwardNeuralNetwork::getF1Score() const;

See more detail on F1 score

Mean Absolute Error

Mainly used for regression.
The MAE is a measure of errors between the output of neural network and the expected output.

Here the function to retreive the weighted MAE:

void StraightforwardNeuralNetwork::getMeanAbsoluteError() const;

See more detail on MAE

Root-Mean-Square Error

RMSE is very similar to MAE except that examples with a large error have a greater impact on the value of RMSE due to the square. Mainly used for regression. It is equivalent to root-mean-square deviation (RMSE).

Here the function to retreive the weighted RMSE:

void StraightforwardNeuralNetwork::getRootMeanSquaredError() const;

See more detail on RMSE