Supersparse linear integer models for optimized medical scoring systems
From MaRDI portal
Publication:506427
DOI10.1007/S10994-015-5528-6zbMATH Open1406.62144arXiv1502.04269OpenAlexW2164878629MaRDI QIDQ506427FDOQ506427
Authors: Berk Ustun, Cynthia Rudin 
Publication date: 31 January 2017
Published in: Machine Learning (Search for Journal in Brave)
Abstract: Scoring systems are linear classification models that only require users to add, subtract and multiply a few small numbers in order to make a prediction. These models are in widespread use by the medical community, but are difficult to learn from data because they need to be accurate and sparse, have coprime integer coefficients, and satisfy multiple operational constraints. We present a new method for creating data-driven scoring systems called a Supersparse Linear Integer Model (SLIM). SLIM scoring systems are built by solving an integer program that directly encodes measures of accuracy (the 0-1 loss) and sparsity (the -seminorm) while restricting coefficients to coprime integers. SLIM can seamlessly incorporate a wide range of operational constraints related to accuracy and sparsity, and can produce highly tailored models without parameter tuning. We provide bounds on the testing and training accuracy of SLIM scoring systems, and present a new data reduction technique that can improve scalability by eliminating a portion of the training data beforehand. Our paper includes results from a collaboration with the Massachusetts General Hospital Sleep Laboratory, where SLIM was used to create a highly tailored scoring system for sleep apnea screening
Full work available at URL: https://arxiv.org/abs/1502.04269
Recommendations
predictioninterpretabilitysparsitysupervised classificationinteger programmingsleep apnea0-1 lossdiscrete linear classificationmedical scoring systems
Classification and discrimination; cluster analysis (statistical aspects) (62H30) Applications of statistics to biology and medical sciences; meta analysis (62P10) Applications of mathematical programming (90C90)
Cites Work
- A lasso for hierarchical interactions
- Least angle regression. (With discussion)
- 10.1162/153244303322753616
- Title not available (Why is that?)
- Regularization and Variable Selection Via the Elastic Net
- Title not available (Why is that?)
- Scenario reduction in stochastic programming
- Title not available (Why is that?)
- Structured variable selection with sparsity-inducing norms
- Advanced Lectures on Machine Learning
- Breast Cancer Diagnosis and Prognosis Via Linear Programming
- Scenarios for multistage stochastic programs
- Integer programming methods for normalisation and variable selection in mathematical programming discriminant analysis models
- Misclassification minimization
- Binarized support vector machines
- Support Vector Machines with the Ramp Loss and the Hard Margin Loss
- Mathematical Programming for Data Mining: Formulations and Challenges
- Analysis of the consistency of a mixed integer programming-based multi-category constrained discriminant model
- Sparse weighted voting classifier selection and its linear programming relaxations
- Selection of Subsets of Regression Variables
- Solving mixed integer classification problems by decomposition
- Mathematical programming formulations for two-group classification with binary variables
- A combinatorial approach to the classification problem
- Fine-scale statistics for the multidimensional Farey sequence
- Integer Programming Solution of a Classification Problem
Cited In (32)
- Multiple criteria sorting models and methods. I: Survey of the literature
- Title not available (Why is that?)
- On mathematical optimization for clustering categories in contingency tables
- Visualizing the Implicit Model Selection Tradeoff
- Mathematical optimization in classification and regression trees
- Learning decision catalogues for situated decision making: the case of scoring systems
- Learning certifiably optimal rule lists for categorical data
- Optimal randomized classification trees
- A semidefinite programming method for integer convex quadratic minimization
- Title not available (Why is that?)
- Title not available (Why is that?)
- A framework for inherently interpretable optimization models
- Title not available (Why is that?)
- Sparsity in optimal randomized classification trees
- High dimensional data classification and feature selection using support vector machines
- The backbone method for ultra-high dimensional sparse machine learning
- Mixed integer quadratic optimization formulations for eliminating multicollinearity based on variance inflation factor
- How to find a good explanation for clustering?
- A survey on the explainability of supervised machine learning
- Investigating consumers' store-choice behavior via hierarchical variable selection
- Interpretable machine learning: fundamental principles and 10 grand challenges
- Explainable and interpretable machine learning and data mining
- Quadratic Convergence of Smoothing Newton's Method for 0/1 Loss Optimization
- Learning optimized risk scores
- Disjunctive Rule Lists
- Best subset selection via cross-validation criterion
- A decision-theoretic approach for model interpretability in Bayesian framework
- Learning non-compensatory sorting models using efficient SAT/MaxSAT formulations
- Algorithms for the unit-cost stochastic score classification problem
- Learning customized and optimized lists of rules with mathematical programming
- On sparse optimal regression trees
- Sparse optimization via vector \(k\)-norm and DC programming with an application to feature selection for support vector machines
Uses Software
This page was built for publication: Supersparse linear integer models for optimized medical scoring systems
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q506427)
