All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Added wheel package supported for
py3.12
- Update
pandas
version to>=1.5
- Python
3.7
is no longer supported
- Added
performance
intoad.utils
for test code performance - Added
pickletracer
intoad.utils
for infer requirements in pickle object
- Fixed
Value Error
inselect
anddrop_corr
method when usingpandas >= 2.0.x
- Added
ks_plot
for KS plot, #102 thanks @kevin-meng - Added
xgb_loss
decorator for convert a normal loss function to a xgb supported loss function - Added
binary_focal_loss
function innn.functional
- Added
event
module innn.trainer
, and changedtrainer
mode to event-based - Added wheel package supported for
py3.9
,py3.10
andpy3.11
- Now you can pass arguments to
DecisionTreeClassifier
inmerge
orCombiner
when usemethod = dt
- Fixed
groupby
rewrited inpreprocessing
- Fixed the expired deprecations of numpy types in
1.24.0
- Added
Progress
forpandas.apply
by usingpandas_enable
andpandas_disable
- Added
feature_bin_stats
for feature bins, #91 thanks @kevin-meng
countBlank
can use customize missing value, #101 thanks @kevin-meng- remove ref of
merge
in__init__
file
0.1.0 - 2021-10-08
- Added
backward_rounds
fornn.Trainer.train
- Added
evalute
func innn.Module
- Added
get_reason
func inScoreCard
, #79 thanks @qianweishuo - Added dict type input support for
ScoreCard.predict
andCombiner.transform
, #79 thanks @qianweishuo - Added iterator support for
Progress
- Change
callback
andearlystopping
to python decorator
0.0.65 - 2021-06-30
- Add new
lift
value and rename the oldlift
value tocum_lift
inKS_Bucket
- Move
nn.autoencoder
tonn.zoo.autoencoder
- Added
label_smoothing
,focal_loss
function innn
module - Added some features in
nn.trainer
- Added default
early_stopping
fornn.Trainer
- Update
numpy
version to>=1.20
- Python
3.6
is no longer supported
- Fixed combiner error after
ScoreCard
reload. #67
0.0.64 - 2021-03-22
- Added
callback
param infit
method fornn
- Added
Trainer
andEarlyStopping
innn.trainer
module
- Use mean of loss in
nn.Module.fit
instead of the latest loss value - Set default rotation for x tick labels
- Fixed dependence version of
numpy
- Fixed
DistModule
module - Fixed
ScoreCard
representation error
0.0.62 - 2021-02-19
save
andload
method for nn module- Added
lift
value inKS_bucket
function - Added checking duplicate keys in
Transformer
quality
method supportindicators
- Fixed tadpole warning of legend. #52
- Fixed tadpole
title
andx/y label
display forUTF8
- Fixed default rule in RuleMixin.
- Fixed loss function of VAE model.
- Fixed
decimal
argument inScoreCard.export
function
- Reduce memory usage when using
select
function
0.0.61 - 2020-06-24
- Support for calculating IV for each groups in a feature. #25
- Add
cpu_cores
forquality
function - Add
predict_proba
forScoreCard
- Impute module
- NN module
- The y axis of
badrate_plot
is starting with0
now. #23 KS
is implemented usingks2samp
instead
- Fixed
Preprocess
bugs
- Add references for
Chi-Merge
,Stepwise Regression
,Scorecard Transformation
0.0.60 - 2020-04-20
- Preprocess module.
- Annotation format for bin plot.
- KS bucket support split pointers as bucket. #22
- Format_bins support ellipsis.
- Reverse cumulative columns in KS bucket
- Use correct order of score for auc and roc plot. #21
- Fixed number type of x axis of badrate plot. #20
- Fixed negative ks value in
KS_bucket
.
0.0.59 - 2020-02-07
- Combiner support empty separate.
- Confusion matrix function in metrics.
- support python 3.8.
- Transform support y as string type.
- VIF independent statsmodels.