drsteve / pyforecasttools Goto Github PK
View Code? Open in Web Editor NEWForecast Verification/Validation Tools in Python
License: BSD 3-Clause "New" or "Revised" License
Forecast Verification/Validation Tools in Python
License: BSD 3-Clause "New" or "Revised" License
Feature Request
The dictionary returned by plot.rocCurve
has Figure
, Axes
, POD
, POFD
, and Thresholds
.
It'd be extremely useful to add the area under the curve (AUC) as a measure of model performance.
Currently the implementation uses Contingency2x2.fromBoolean()
and loops over thresholds creating a new Contingency2x2
object each time. It might be good to consider directly supporting probabilistic predictions, possibly in a similar manner to fromBoolean
- this could them allow updating of the contingency table by updating the threshold as the probabilities would be stored with the object (and would allow bootstrapping CIs, same as fromBoolean
).
AUC could then be calculated without the ROC plotting.
The reliability diagram (in the plot
module) currently automatically determines the number of bins to use for the reliabilty diagram by the Freedman-Diaconis method, with a hard upper limit on the number of bins (for graphical purposes).
For datasets with limited numbers of events (i.e., imbalanced datasets) this can lead to bin sizes being much smaller than desired. It would be useful to allow specification of the number of bins, or bin selection method, in the call to verify.plot.reliabilityDiagram
.
Current documentation is sparse. It would be helpful to users to provide:
Installation produces 5 files: dependency_links.txt, PKG-INFO, requires.txt, SOURCES.txt, and top_level.txt. While I'm not sure why all of these files are needed (especially top_level.txt), I think they should be added to .gitignore.
Feature Request
Currently the plot.reliabilityDiagram
feature plots the predicted probability against empirical probability, as well as adding a histogram of the predicted probabilities.
It'd be extremely useful have options to add:
The documentation is ugly. The docstring formatting isn't presently being rendered correctly by Sphinx.
The workflow for the github.io webpage for this project currently follows https://daler.github.io/sphinxdoc-test/includeme.html
ETA: Documentation is at https://drsteve.github.io/PyForecastTools, and auto-updates whenever the gh-pages branch of this repo is updated.
Missing unit tests for the functions.
Steve,
minor warning that is likely not important but thought I should put it here.
numpy=1.18.3
verify.metrics.medSymAccuracy(fitted_model(x), data)
/Users/balarsen/miniconda3/envs/python3/lib/python3.7/site-packages/numpy/core/fromnumeric.py:746: UserWarning: Warning: 'partition' will ignore the 'mask' of the MaskedArray.
a.partition(kth, axis=axis, kind=kind, order=order)
After installing from pip, some of the functions such as bias
, etc, are not attached to verify
when importing. After uninstalling from pip and reinstalling from github, all functions are importing correctly. I suspect that switching to absolute import statements in init
would fix this on the pip side.
from verify.metrics import *
from verify.categorical import *
from verify. import plot
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.