Releases: oegedijk/rule_estimator
Releases · oegedijk/rule_estimator
v0.4.1 bump dash requirement for dcc.Clipboard support
v0.4.0: improved dashboard, some breaking changes
Delete dashboard_demo.mp4
v0.3.0: introducing dashboard
This release adds a dashboard functionality based on plotly dash that allows you to interactively build decision rules using parallel plots of your data.
There are many, many breaking changes, so almost all 0.2.*
code will no longer work.
v0.2.1: parallel_coordinates and MultiRange rules
Version 0.2.1:
New Features
- new
estimator.parallel_coordinates(X, y, rule_id)
method to plot a parallel
coordinates plot of data entering rulerule_id
. - new rules:
MultiRangeAndRule
andMultiRangeOrRule
. - new nodes:
MultiRangeAndNode
andMultiRangeOrNode
.
Bug Fixes
- Fixes bugs with
replace_rule
andappend_rule
Improvements
append_rule
now also inserts in the correct position whenrule_id
is
inside aCaseWhen
rule
v0.2: edit and score your rules
version 0.2:
Breaking Changes
- Custom rules are now defined with
__rule__
method that returns a boolean mask
instead of withpredict(X)
method. DummyRule
is now calledPredictionRule
New Features
- each rule now gets assigned a
rule_id
, which is displayed when you call
estimator.describe()
- new
score_rules(X, y)
method that shows performance of individual rules - new
get_igraph()
method, that returns an igraph Graph object of the rules - new
plot()
method that returns a plotly figure of the rules - new
get_rule(rule_id)
,replace_rule(rule_id, new_rule)
andappend_rule(rule_id, new_rule)
methods - new
get_rule_params(rule_id)
andset_rule_params(rule_id, **params)
methods - new
get_rule_input(rule_id, X, y)
andget_rule_leftover(rule_id, X, y)
to get the specific data
that either flows into a rule, or the unlabeled data that flows out of a rule.
This helps in constructing new rules as you can target it to the data
that would appear in that part of the rule graph.
Improvements
- data is now split up and only non-labeled data is passed to downstream rules.
v0.1.2: using find_packages() in setup.py
Merge branch 'master' of https://github.com/oegedijk/rule_estimator i…
v0.1.1: setup.py packages
Update setup.py
v0.1: initial release
Initial release!
Features:
- both RuleClassifier and RuleRegressor
- save to and load from yaml files
- create your own business rules or use predefined ones