Gin077的专栏

全局代理和局部代理一、Global Surrogate全局代理具体步骤如下：Perform the following steps to obtain a surrogate model:1. Select a dataset X. This can be the same dataset that was used for training the black box model...

一、Feature InteractionThe interaction between two features is the change in the prediction that occurs by varying the features after considering the individual feature effects.二、Feature Importance...

一、作为模型代理方法的第一节，先介绍模型代理方法的思路从world捕捉data，用data训练模型，再用可解释性方法来对模型的结果给出解释。把模型训练和模型解释分开，使得训练模型不再局限在拥有内在可解释性的模型范围内。二、下面介绍第一种Partial Dependence Plot (PDP)，部分依赖图形。描述的是单个或两个feature对模型outcome的边际影响。The...

一、引入Q：The linear regression model does not account for interactions between features. Would it not be convenient to have a model that is as simple and interpretable as linear models, but also integr...

可以说是最简单的model了，IF-THEN的结构一、The usefulness of a decision rule is usually summarized in two numbers: Support and accuracy。Support（coverage of a rule）：走进该rule的实例占比，The percentage of instances to whic...

这本书以CART树为例一、CART树分裂节点的过程1、对于回归问题，最小化y的方差来决定分裂点。The variance tells us how much the y values in a node are spread around their mean value2、对于分类问题，最小化y的GINI系数，The Gini index tells us how “impure” ...

一、Logistic Regression前面的可参考linear regression，Logistics Regression只是在后面加了logistic function（1）logistic function（2）Interpretation二、广义模型线性模型中的三个假设在现实中无法满足Three assumptions of the linear ...

关于模型可解释性涉及到模型的整体可解释性和单个实例的可解释性。这里着重强调的是实例级别（instance-level）的可解释性。解释方法分为自带可解释性的模型和模型无关的方法。本节主要介绍自带可解释性的模型。主要包含：Monotone：是否具有单调性，即feature和target是否单调Interation：是否自带特征交叉Linear Regressiony = β0...