Metrics

Functions:

plot_confusion_matrix(true_labels, …[, …])	Generates a confusion matrix for the given predictions.
precision_recall_f1(true_labels, …)	Generates classification metrics precision, recall and F1 for the given predictions.
save_experiment(file_name, params, …)	Saves/appends all experiment parameters and results to .csv file.
save_history(file_name, history)	Saves training history dictionary as numpy arrays in .npz file.
save_predictions(file_name, true_labels, …)	Saves predictions to .csv file.
save_results(file_name, test_loss, …)	Saves test metrics to a .txt file.

metrics.plot_confusion_matrix(true_labels, predicted_labels, labels, matrix_dim=15, normalise=False, title=None, fig_size=(10, 10), font_size=15)

Generates a confusion matrix for the given predictions.

Uses sklearn to generate the confusion matrix. Uses matplotlib and seaborn to generate the figure.

Parameters

• true_labels (np.array) – The ground truth labels for the data

• predicted_labels (np.array) – The predicted labels for the data

• labels (list) – List of label names

• title (str) – The title for the figure

• matrix_dim (int) – The number of classes to show on the matrix, if None or -1 creates full matrix

• normalise (bool) – Whether to normalise the matrix or use original values

• fig_size (tuple) – Tuple for the horizontal and vertical size the figure

• font_size (int) – Font size for figure labels.

Returns

The confusion matrix figure matrix (numpy array): The 2d confusion matrix array

Return type

fig (matplotlib figure)

metrics.precision_recall_f1(true_labels, predicted_labels, labels)

Generates classification metrics precision, recall and F1 for the given predictions.

Parameters

• true_labels (np.array) – The ground truth labels for the data

• predicted_labels (np.array) – The predicted labels for the data

• labels (list) – List of label names

Returns

Dictionary of micro, macro and weighted precision, recall and F1 scores cls_report_str (str): Formatted string with metric data for all classes and totals cls_report_dct (dict): Dictionary containing all metric data of metrics string

Return type

metrics_dict (dict)

metrics.save_experiment(file_name, params, train_loss, train_acc, val_loss, val_acc, test_loss, test_acc, metrics)

Saves/appends all experiment parameters and results to .csv file.

metrics.save_history(file_name, history)

Saves training history dictionary as numpy arrays in .npz file.

metrics.save_predictions(file_name, true_labels, predicted_labels, predictions)

Saves predictions to .csv file.

metrics.save_results(file_name, test_loss, test_accuracy, metrics)

Saves test metrics to a .txt file.