Learning optimized risk scores
From MaRDI portal
Abstract: Risk scores are simple classification models that let users make quick risk predictions by adding and subtracting a few small numbers. These models are widely used in medicine and criminal justice, but are difficult to learn from data because they need to be calibrated, sparse, use small integer coefficients, and obey application-specific operational constraints. In this paper, we present a new machine learning approach to learn risk scores. We formulate the risk score problem as a mixed integer nonlinear program, and present a cutting plane algorithm for non-convex settings to efficiently recover its optimal solution. We improve our algorithm with specialized techniques to generate feasible solutions, narrow the optimality gap, and reduce data-related computation. Our approach can fit risk scores in a way that scales linearly in the number of samples, provides a certificate of optimality, and obeys real-world constraints without parameter tuning or post-processing. We benchmark the performance benefits of this approach through an extensive set of numerical experiments, comparing to risk scores built using heuristic approaches. We also discuss its practical benefits through a real-world application where we build a customized risk score for ICU seizure prediction in collaboration with the Massachusetts General Hospital.
Recommendations
- Supersparse linear integer models for optimized medical scoring systems
- Support vector hazards machine: a counting process framework for learning risk scores for censored outcomes
- Learning Theory
- The Effectiveness of Risk Scores: The Logit Rank Plot
- Model-free scoring system for risk prediction with application to hepatocellular carcinoma study
Cites work
- A Bayesian framework for learning rule sets for interpretable classification
- A semidefinite programming method for integer convex quadratic minimization
- Algorithms and Software for Convex Mixed Integer Nonlinear Programs
- An interior-point Benders based branch-and-cut algorithm for mixed integer programs
- Best subset selection via a modern optimization lens
- Breast Cancer Diagnosis and Prognosis Via Linear Programming
- Bundle methods for regularized risk minimization
- Combinatorial Benders cuts for the minimum tollbooth problem
- Composite binary losses
- Concentration inequalities for sampling without replacement
- Cutting-plane training of structural SVMs
- Feature subset selection for logistic regression via mixed integer optimization
- scientific article; zbMATH DE number 6617274 (Why is no real title available?)
- scientific article; zbMATH DE number 2107836 (Why is no real title available?)
- scientific article; zbMATH DE number 7064055 (Why is no real title available?)
- scientific article; zbMATH DE number 5066287 (Why is no real title available?)
- Interpretable classifiers using rules and Bayesian analysis: building a better stroke prediction model
- Learning certifiably optimal rule lists for categorical data
- Learning customized and optimized lists of rules with mathematical programming
- Learning decision trees with flexible constraints and objectives using integer optimization
- Learning optimized risk scores
- On equivalence relationships between classification and ranking algorithms
- On over-fitting in model selection and subsequent selection bias in performance evaluation
- Optimized cutting plane algorithm for large-scale risk minimization
- Piecewise-linear approximation for feature subset selection in a sequential logit model
- Polyhedral approximation in mixed-integer convex optimization
- Sparse weighted voting classifier selection and its linear programming relaxations
- Supersparse linear integer models for optimized medical scoring systems
- The Cutting-Plane Method for Solving Convex Programs
- Very simple classification rules perform well on most commonly used datasets
Cited in
(15)- Sparse learning of the disease severity score for high-dimensional data
- Interpretable machine learning: fundamental principles and 10 grand challenges
- Algorithms for the unit-cost stochastic score classification problem
- Metarisk: semi-supervised few-shot operational risk classification in banking industry
- Multiple criteria sorting models and methods. I: Survey of the literature
- The stochastic score classification problem
- Supersparse linear integer models for optimized medical scoring systems
- A survey on the explainability of supervised machine learning
- Learning optimized risk scores
- Learning decision catalogues for situated decision making: the case of scoring systems
- Explainable and interpretable machine learning and data mining
- Rank-one convexification for sparse regression
- Probabilistic scoring lists for interpretable machine learning
- A guided tour of multiple criteria sorting models and methods
- Towards a trade-off of interpretability, accuracy and scalability: enhanced formulations in linear classification models
This page was built for publication: Learning optimized risk scores
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5214243)
