Reference¶
BFASTMonitor Implementation¶
-
class
bfast.BFASTMonitor(start_monitor, freq=365, k=3, hfrac=0.25, trend=True, level=0.05, period=10, verbose=0, backend='opencl', platform_id=0, device_id=0, detailed_results=False, find_magnitudes=True)¶ BFASTMonitor implements the BFASTMonitor approach and provides two backends/implementations:
A pure Python implementation (based on the Numpy package).
An optimized OpenCL implementation suited for massively-parallel devices.
The interface follows the one of the corresponding R package, see: https://cran.r-project.org/web/packages/bfast
- Parameters
- start_monitordatetime
A datetime object specifying the start of the monitoring phase.
- freqint, default 365
The frequency for the seasonal model.
- kint, default 3
The number of harmonic terms.
- hfracfloat, default 0.25
Float in the interval (0,1) specifying the bandwidth relative to the sample size in the MOSUM/ME monitoring processes.
- trendbool, default True
Whether a tend offset term shall be used or not.
- levelfloat, default 0.05
Significance level of the monitoring (and ROC, if selected) procedure, i.e., probability of type I error.
- verboseint, optional (default=0)
The verbosity level (0=no output, 1=output).
- backendstr, default ‘opencl”
Specifies the implementation that shall be used: backend=’python’ resorts to the non-optimized Python version; backend=’opencl’ resorts to the optimized massively-parallel OpenCL implementation.
- platform_idint, default 0
Only relevant if backend=’opencl’. Specifies the OpenCL platform id.
- device_id int, default 0
Only relevant if backend=’opencl’. Specified the OpenCL device id.
- detailed_resultsbool, default False
Only relevant if backend=’opencl’. If detailed results should be returned or not.
- find_magnitudesbool, default True
Only relevant if backend=’opencl’. If magnitudes should be returned or not. Disabling this would improve the performance greatly
- Attributes
-
fit(self, data, dates, n_chunks=None, nan_value=0)¶ Fits the models for the ndarray ‘data’
- Parameters
- data: ndarray of shape (N, W, H),
where N is the number of time series points per pixel and W and H the width and the height of the image, respectively.
- dateslist of datetime objects
Specifies the dates of the elements in data indexed by the first axis n_chunks : int or None, default None
- nan_valueint, default 0
Specified the NaN value used in the array data
- Returns
- selfThe BFASTMonitor object.
-
get_params(self)¶ Returns the parameters for this model.
- Returns
- dict: Mapping of string to any
parameter names mapped to their values.
-
set_params(self, **params)¶ Sets the parameters for this model.
- Parameters
- paramsdict
Dictionary containing the parameters to be used.
-
property
timers¶ Returns runtime measurements for the different phases of the fitting process.
- Returns
- dictAn dictionary containing the runtimes
for the different phases.
-
property
breaks¶ Returns the breaks that have been computed.
- Returns
- array-like: An array containing the (first) breaks
detected by BFASTMonitor. A ‘-2’ corresponds to pixels (time series) not containing sufficient data and a ‘-1’ to pixels without any breaks. All other non-negative entries correspond to the first break that was detected in the monitor period (i.e., its index).
-
property
means¶ Returns the means computed
- Returns
- array-likeAn array containing the mean values
of the individual MOSUM processes (e.g., a positive mean for a pixel corresponds to an increase of the vegetation in case indices such as NDMI are considered)
-
property
magnitudes¶ Returns the magnitudes computed
- Returns
- array-likeAn array containing the magnitude
values median of the difference between the data and the model prediction in the monitoring period
-
property
valids¶ Returns the number of valid values for each pixel
- Returns
- array-likeAn array containing the number
of valid values for each pixel in the aray data
Helper Functions¶
-
bfast.monitor.utils.crop_data_dates(data, dates, start, end)¶ Crops the input data and the associated dates w.r.t. the provided start and end datetime object.
- Parameters
- data: ndarray of shape (N, W, H)
Here, N is the number of time series points per pixel and W and H are the width and the height of the image, respectively.
- dateslist of datetime objects
Specifies the dates of the elements in data indexed by the first axis n_chunks : int or None, default None
- startdatetime
The start datetime object
- enddatetime
The end datetime object
- Returns
- Returns: data, dates
The cropped data array and the cropped list. Only those images and dates that are with the start/end period are contained in the returned objects.