API Reference

The features library provides common features for machine learning models of behavioral modes from acceleration data, as well as a mechanism to extend and add additional features.

pyecoacc.features.transform

ACCStatsTransformer(single_ax_stat_list='all', multi_ax_stat_list='all')

Bases: BaseEstimator, TransformerMixin

sklearn style Transformer to computes the statistics for each ACC sample. This is then used as a first step in Pipeline objects. The input data is expected to be a 2D numpy array of shape (n_samples, n_timesteps*3) where each sample contains interleaved x, y, z accelerometer data. The output is a pandas DataFrame where each row corresponds to a sample and each column to a computed feature.

Features computed include both single-axis statistics (like mean, std, etc.) for each axis (x, y, z) and the norm, as well as multi-axis statistics (like correlation between axes).

Parameters:

Name	Type	Description	Default
single_ax_stat_list	str	The features to use on each axis. Defaults to "all".	'all'
multi_ax_stat_list	str	The multiple-axis features to use. Defaults to "all".	'all'

Source code in pyecoacc/features/transform.py

def __init__(self, single_ax_stat_list="all", multi_ax_stat_list="all"):
    """

    Args:
        single_ax_stat_list (str, optional): The features to use on each axis. Defaults to "all".
        multi_ax_stat_list (str, optional): The multiple-axis features to use. Defaults to "all".
    """
    self.single_ax_stat_list = single_ax_stat_list
    self.multi_ax_stat_list = multi_ax_stat_list

pyecoacc.features.stats

register_multiple_axis_feature(name, func)

Add a multiple-axis feature to the feature registry.

Parameters:

Name	Type	Description	Default
name	str	the name of the added feature.	required
func	function	the function that computes the feature. The function should take four 2D numpy arrays (X, Y, Z, norm) and return a 1D numpy array (samples,).	required

Source code in pyecoacc/features/stats.py

def register_multiple_axis_feature(name, func):
    """Add a multiple-axis feature to the feature registry.

    Args:
        name (str): the name of the added feature. 
        func (function): the function that computes the feature. The function should take four 2D numpy arrays (X, Y, Z, norm) and return a 1D numpy array (samples,).
    """
    multiple_axis_features[name] = func 

register_single_axis_feature(name, func)

Add a single-axis feature to the feature registry.

Parameters:

Name	Type	Description	Default
name	str	the name of the added feature.	required
func	function	the function that computes the feature. The function should take a 2D numpy array (samples x timepoints) and return a 1D numpy array (samples,).	required

Source code in pyecoacc/features/stats.py

def register_single_axis_feature(name, func):
    """Add a single-axis feature to the feature registry.

    Args:
        name (str): the name of the added feature. 
        func (function): the function that computes the feature. The function should take a 2D numpy array (samples x timepoints) and return a 1D numpy array (samples,).
    """
    single_axis_features[name] = func

set_ODBA_low_pass_window(window_size)

Set the smoothing window size for the ODBA calculations.

Parameters:

Name	Type	Description	Default
window_size	int	the number of samples to use in the moving average filter.	required

Source code in pyecoacc/features/stats.py

def set_ODBA_low_pass_window(window_size):
    """Set the smoothing window size for the ODBA calculations.

    Args:
        window_size (int): the number of samples to use in the moving average filter.
    """
    global ODBA_LOW_PASS_WINDOW, window
    ODBA_LOW_PASS_WINDOW = window_size
    window = np.ones(ODBA_LOW_PASS_WINDOW) / ODBA_LOW_PASS_WINDOW