An Interpretable Stroke Prediction Model using Rules and Bayesian Analysis

By Benjamin Letham, Cynthia Rudin, Tyler H. McCormick, David Madigan. 2013

53 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Medical & Health AI

Citations in the corpus

"Wrong side of the tracks": Big Data and Protected Categories (2014, background): “[11] are two recent examples, that allow us to see how the different parts of an al gorithm’s input are combined to produce the final prediction.”
Learning Optimized Risk Scores (2016, methodology): “ referred to as the \weighted Burgess" method. Risk scores built using ad hoc approaches can perform poorly compared to models t using modern methods (see e.g., the poor performance of CHADS2 in Letham et al. 2013, and of risk scores built using the Burgess method in Duwe and Kim 2016). In ligh”
Supersparse Linear Integer Models for Predictive Scoring Systems (2013, methodology): “Existing approaches to interpretable modeling include decision trees and lists (Rüping 2006; Quinlan 1986; Rivest 1987; Letham et al. 2013).”
Methods and Models for Interpretable Linear Classification (2014, background): “Popular transparent classification models include linear models (addressed in this work), decision trees [88, 110, 90], decision lists [94, 65], and decision tables [58].”
Explainable Artificial Intelligence: a Systematic Review (2020, methodology): “The Bayesian Rule Sets (BRS) method [276, 277] is similar to BRL but it uses different probabilities, with the posterior as a Bernoulli distribution, and the prior a Beta distribution whose parameters can be adjusted by end-users to guide BRS toward more interpretable solutions by specifying the des”
Interpretable Policies for Dynamic Product Recommendations (2016, background): “Lack of interpretable solutions can be an important road-block in many critical domains, such as medicine [12, 11, 26].”
Mining Electronic Health Records: A Survey (2017, background): “…[Gotz et al. 2011a; Vellanki et al. 2014; Miotto et al. 2016] Descritive Temporal Cross-Sectional [Wilcox and Hripcsak 2003; Sarkar et al. 2012; Letham et al. 2013; Ebadol-lahi et al. 2010; Feldman and Chawla 2014; Stiglic et al. 2015; Ngufor et al. 2015; Byrd et al. 2014; Liang et al. 2014]…”
Interpretability of machine learning models and representations: an introduction (2016, background): “For instance, two decision rule lists/sets can be compared in terms of their number of rules and terms [21,22] and two decision trees can be compared in terms of their number of nodes [23].”
Analysis of Explainers of Black Box Deep Neural Networks for Computer Vision: A Survey (2019, methodology): “The work was developed from preliminary versions that used a different prior, called Bayesian List Machine [68].”
Machine learning for science and society (2013, background): “…us include criminol-ogy (Wang et al. 2013), crop yield prediction (Wagstaff et al. 2008), the energy grid (Rudin et al. 2010, 2012), healthcare (Letham et al. 2013b; McCormick et al. 2012), information retrieval (Letham et al. 2013a), interpretable models (Letham et al. 2013b; McCormick et al.…”
On Explaining Decision Trees (2020, background): “Motivated by this perception, there has been extensive work on learning DTs (and related logic ML models) with properties deemed important for interpretability, e.g. number of nodes, maximum/average depth, etc. [2,68,25,32,1,7,57,24,26,70,67,58,44,30,17,5,73,35,4,78,69,10,34,72,71,36,46,11,45] 2 .”
Learning theory analysis for association rules and sequential event prediction (2013, methodology): “The work of Letham et al. (2013b) uses empirical risk minimization to solve sequential event prediction problems dealing with email recipient recommendation, healthcare, and cooking.”
Supersparse Linear Integer Models for Interpretable Classification (2013, methodology): “an CHADS 2, even though it is a more complicated model. It has also been shown that there exist predictive models that are just as parsimonious and interpretable as CHADS 2, but have better accuracy [Letham et al., 2013]. This demonstrates the dangers of constructing predictive models by hand. The S”