Python Chefboost功能重要性未找到类似outputs/rules/rules_fi.csv的文件

Question

我正在使用Chefboost构建Chaid决策树，并希望检查特征的重要性。出于某种原因，我出现了以下错误：

cb.feature_importance() Feature importance calculation is enabled when parallelised fitting. It seems that fit function didn't called parallelised. No file found like outputs/rules/rules_fi.csv

这是我的代码：

from chefboost import Chefboost as cb
X_train['Decision']=y_train
config={'algorithm': 'CHAID','enableParallelism': enableParallelism}
cb.fit(X_train,config)
cb.feature_importance()

有人能帮我吗？ 谢谢

Answer 1

你不必再并行运行了。特征重要性函数需要rules.py的精确路径。一定要先升级你的chefboost库

config={'algorithm': 'CHAID'}
model = cb.fit(X_train, config)

#get decision rules
#decision rules = "outputs/rules/rules.py" #static way
decision_rules = model["trees"][0].__dict__["__spec__"].origin #dynamic way
cb.feature_importance(decision_rules)

Answer 2

通过使用下面的代码解决了该问题

from chefboost import Chefboost as cb
X_train['Decision']=y_train
config={'algorithm': 'CHAID','enableParallelism': True}
cb.fit(X_train,config) 
cb.feature_importance()