Evaluation

Scripts to perform evaluation on the data

Auxiliar Functions

plot_metric_heatmaps

 plot_metric_heatmaps (results:dict, distance_metrics:list,
                       clustering_algorithms:list,
                       evaluation_metrics:list)

Plot heatmaps for each evaluation metric showing scores across different distance metrics and clustering algorithms combinations.

	Type	Details
results	dict	Dictionary containing results for each combination of distance metric and clustering algorithm
distance_metrics	list	List of distance metric names
clustering_algorithms	list	List of clustering algorithm names
evaluation_metrics	list	List of evaluation metric names
Returns	None

source

find_non_matching_elements

 find_non_matching_elements (main_array:numpy.ndarray,
                             check_array:numpy.ndarray)

Finds elements in check_array that are not present in main_array.

	Type	Details
main_array	ndarray	The main array with larger set of elements
check_array	ndarray	The array with elements to check against the main array
Returns	ndarray

Evaluate Clustering with Multiple Labels

source

evaluate_clustering_multiple_labels

 evaluate_clustering_multiple_labels
                                      (latent_representations:numpy.ndarra
                                      y, list_of_labels:list,
                                      clustering_method:str='kmeans',
                                      label_names:list=None, **kwargs)

Evaluates the clustering quality of the latent representations for one or multiple sets of labels.

	Type	Default	Details
latent_representations	ndarray		The latent space data.
list_of_labels	list		List of true labels or a single true labels array.
clustering_method	str	kmeans	The clustering algorithm to use (‘kmeans’, ‘gmm’, ‘dbscan’).
label_names	list	None	Optional names for the label sets.
kwargs	VAR_KEYWORD
Returns	dict		Returns a dictionary with clustering metrics.

Physical Distances

Euclidean

source

euclidean_distance

 euclidean_distance (point1:numpy.ndarray, point2:numpy.ndarray)

Manhattan

source

manhattan_distance

 manhattan_distance (point1:numpy.ndarray, point2:numpy.ndarray)

Cosine

source

cosine_distance

 cosine_distance (point1:numpy.ndarray, point2:numpy.ndarray)

Dynamic Time Warping

source

dtw_distance

 dtw_distance (point1:numpy.ndarray, point2:numpy.ndarray)

Generic

source

calculate_distance

 calculate_distance (point1:numpy.ndarray, point2:numpy.ndarray,
                     distance_metric:str='euclidean')

*Calculates the distance between two points based on the specified distance metric.

:param point1: First data point array. :param point2: Second data point array. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). :return: Distance as a float.*

source

calculate_pairwise_distances

 calculate_pairwise_distances (array1:numpy.ndarray, array2:numpy.ndarray,
                               distance_metric:str='euclidean')

*Calculates the distance between corresponding pairs of points from two arrays using the specified distance metric.

:param array1: A 2D numpy array where each row represents a data point. :param array2: A 2D numpy array where each row represents a data point. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). :return: A 1D numpy array containing the distances between corresponding pairs.*

Batch

source

calculate_distances_batch

 calculate_distances_batch (single_points:numpy.ndarray,
                            points_array:numpy.ndarray,
                            distance_metric:str='euclidean')

*Calculates the distances between single data points and an array of data points based on the specified distance metric.

:param single_points: Single data point array or a batch of data points. :param points_array: Array of data points to compare against. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). :return: Array of distances.*

Nearest Points

source

find_nearest_points

 find_nearest_points (single_point:numpy.ndarray,
                      points_array:numpy.ndarray, n:int,
                      distance_metric:str='euclidean')

source

find_nearest_points_batch

 find_nearest_points_batch (single_points:numpy.ndarray,
                            points_array:numpy.ndarray, n:int,
                            distance_metric:str='euclidean')

*Finds the nearest indices and distances for a batch of single data points to an array of data points based on the specified distance metric.

:param single_points: Array of single data points (2D array). :param points_array: Array of data points to compare against (2D array). :param n: Number of nearest points to retrieve for each single point. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). :return: Tuple of nearest indices and nearest distances for each single point.*

Orbits Distance

source

orbits_distances

 orbits_distances (orbit_data1:numpy.ndarray, orbit_data2:numpy.ndarray,
                   distance_metric:str)

*Calculates distances between orbits in two datasets using a specified distance metric.

This function is robust to input shapes. If an input is a 2D array (representing a single orbit), it is automatically converted to a 3D array with one sample. This allows for flexible comparisons between single or multiple orbits.

:param orbit_data1: First set of orbits (shape: [n_samples1, n_features, n_time_steps] or [n_features, n_time_steps]). :param orbit_data2: Second set of orbits or a single orbit. Shape: [n_samples2, n_features, n_time_steps] or [n_features, n_time_steps]. :param distance_metric: A string representing the distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’).

return: NumPy array of distances. - If one input is single and the other is multiple: - Shape: [n_samples1] or [n_samples2] - If both inputs are multiple: - Shape: [n_samples1, n_samples2]*
	Type	Details
orbit_data1	ndarray	Shape: [n_samples1, n_features, n_time_steps] or [n_features, n_time_steps]
orbit_data2	ndarray	Shape: [n_samples2, n_features, n_time_steps] or [n_features, n_time_steps]
distance_metric	str	String representing the distance metric (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’)
Returns	ndarray

Get the Closest Orbits

source

find_nearest_orbits

 find_nearest_orbits (single_orbit:numpy.ndarray,
                      orbit_data:numpy.ndarray, n:int,
                      distance_metric:str='euclidean')

*Finds the n closest orbits in orbit_data to the single_orbit based on the specified distance metric.

:param single_orbit: The reference orbit (shape: [n_features, n_time_steps]). :param orbit_data: The dataset of orbits (shape: [n_samples, n_features, n_time_steps]). :param n: The number of closest orbits to return. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). Defaults to ‘euclidean’. :return: A tuple containing: - Indices of the n closest orbits in orbit_data. - Distances of the n closest orbits.*

source

find_nearest_orbits_batch

 find_nearest_orbits_batch (single_orbits:numpy.ndarray,
                            orbit_data:numpy.ndarray, n:int,
                            distance_metric:str='euclidean')

*Iteratively finds the n closest orbits in orbit_data for each orbit in single_orbits.

param single_orbits: The reference orbits (shape: [num_single_orbits, n_features, n_time_steps]). :param orbit_data: The dataset of orbits to search within (shape: [n_samples, n_features, n_time_steps]). :param n: The number of closest orbits to return for each single_orbit. :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). Defaults to ‘euclidean’. :return: A tuple containing: - A 2D array of shape [num_single_orbits, n] with indices of the n closest orbits. - A 2D array of shape [num_single_orbits, n] with distances of the n closest orbits.*
	Type	Default	Details
single_orbits	ndarray		Shape: [num_single_orbits, n_features, n_time_steps]
orbit_data	ndarray		Shape: [n_samples, n_features, n_time_steps]
n	int		Number of nearest orbits to find
distance_metric	str	euclidean	Distance metric
Returns	tuple

Calculate Pairwise distances

source

calculate_pairwise_orbit_distances

 calculate_pairwise_orbit_distances (orbit_data1:numpy.ndarray,
                                     orbit_data2:numpy.ndarray,
                                     distance_metric:str='euclidean')

*Calculates the distance between corresponding orbits in two orbit datasets.

param orbit_data1: The first set of orbits (shape: [n_samples, n_features, n_time_steps]). :param orbit_data2: The second set of orbits (shape: [n_samples, n_features, n_time_steps]). :param distance_metric: The distance metric to use (‘euclidean’, ‘manhattan’, ‘cosine’, ‘dtw’). Defaults to ‘euclidean’. :return: An array of distances with shape [n_samples].*
	Type	Default	Details
orbit_data1	ndarray		Shape: [n_samples, n_features, n_time_steps]
orbit_data2	ndarray		Shape: [n_samples, n_features, n_time_steps]
distance_metric	str	euclidean	Distance metric
Returns	ndarray

Evaluate Distance Metrics

source

evaluate_distance_metrics_and_clustering

 evaluate_distance_metrics_and_clustering (orbit_data:numpy.ndarray,
                                           true_labels:numpy.ndarray, dist
                                           ance_metrics:Optional[list]=Non
                                           e, clustering_algorithms:Option
                                           al[list]=None, evaluation_metri
                                           cs:Optional[list]=None,
                                           n_clusters:Optional[int]=None,
                                           plot_results:bool=True)

Evaluates combinations of distance metrics and clustering algorithms on orbit data.

	Type	Default	Details
orbit_data	ndarray		Orbit data as multivariate time series [n_samples, n_features, n_time_steps] or point data [n_samples, n_features]
true_labels	ndarray		Array of true labels for the orbit data
distance_metrics	Optional	None	List of distance metrics to use. If None, uses all available metrics
clustering_algorithms	Optional	None	List of clustering algorithms to use. If None, uses all available algorithms
evaluation_metrics	Optional	None	List of evaluation metrics to use. If None, uses all available metrics
n_clusters	Optional	None	Number of clusters. If None, inferred from labels
plot_results	bool	True	If True, plot heatmaps of results
Returns	dict

Machine Learning

source

machine_learning_evaluation

 machine_learning_evaluation (X:numpy.ndarray, y:numpy.ndarray,
                              print_results:bool=False,
                              return_best_model:bool=False,
                              scale_data:bool=True)

Evaluates multiple machine learning algorithms on the provided dataset.

	Type	Default	Details
X	ndarray		Features array, expected to be 2D. Will attempt to reshape if higher dimensions
y	ndarray		Target labels
print_results	bool	False	If True, visualizes the evaluation results
return_best_model	bool	False	If True, returns the best model based on accuracy
scale_data	bool	True	If True, scales the features using StandardScaler
Returns	Union