A tool for change-preserving classification of multi-temporal (spatial) data.
POCC is a method for data classification, which can reduce these effects of change loss and change exaggeration as far as possible. The classification uses a sweep line algorithm, whose optimal solution is determined with the help of a measure called Preservation Of Change Classes (POCC). By assigning weights during computation of this measure, different tasks or change analyses (e.g. emphasize only highly significant changes) can be processed. As a side effect, such constraints also help to reduce the well-known effect of change blindness. (TO DO: cite paper here)
Requirements:
- python 3 (tested with 3.9)
Set up a python environment and run pip install -r requirements.txt.
python pocc.py [-h] [-p P] [--nodata NODATA] [--startcolumn STARTCOLUMN] [--keys KEYS [KEYS ...]] [--vkey VKEY] filename classes
with
- h: show a help message
- filename: path to a csv file with the data
- classes: the number of target classes.
- P: the factor to mark a change significant (TO DO: needs better explanation)
- NODATA: the value of missing data fields. These will be ignored.
- startcolumn: csv only. the column (zero-indexed) in which the timeseries data starts. All columns to the right will be treated as data.
- keys: geojson only. Name of the properties elements that contain data values. E.g.
--keys 2005 2010 2015 - vkey: geojson only. Name of the properties element that contains a list of data. E.g.
--vkey datawhen your geojson contains properties like "data": ["533.9","494.4","516.5","531.2"]
For example, if you want to classify the sample data into 4 classes, run:
python pocc.py data/sampledata_cars_2005-2020.csv 4 --startcolumn 2
or
python pocc.py data/sampledata_cars_2005-2020.geojson 4 --vkey data
or
python pocc.py data/sampledata_cars_2005-2020_b.geojson 4 --keys 2005 2010 2015 2020
When providing a geojson as input, png files with visualisation of the different epochs will be created in $PWD.