Bump hunting in high-dimensional data

By Jerome H. Friedman, Nicholas I. Fisher. 1999

How researchers see this paper

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

Citations in the corpus

Boolean Decision Rules via Column Generation (2018, background): “Predominant practical approaches include a covering or separate-and-conquer strategy ([15, 14, 16, 26, 28, 40],”
Box drawings for learning with imbalanced data (2014, background): “Some data sets (castle, corner, diamond, square, ﬂooded, castle3D, corner3D, diamond3D, ﬂooded3D, ﬂooded3D) are simulated data that are able to be visualized (made publicly available at (Goh and Rudin 2014)).”
Or's of And's for Interpretable Classification, with Application to Context-Aware Recommender Systems (2015, methodology): “This form of model has some recent precedent in [16, 18, 19, 28] as a form of model that is natural for modeling consumer behavior, and interpretable to human experts.”
Continuous latent variable models for dimensionality reduction and sequential data reconstruction (2001, methodology): “PRIM (Friedman and Fisher, 1999); or some nonparametric and parametric clustering methods, e.”
Explanation Mining: Post Hoc Interpretability of Latent Factor Models for Recommendation Systems (2018, background): “Association rule models have been described as a natural representation of how humans make decisions [35] given their ‘and’, ‘or’ structure and have been succesfully applied to modelling user behaviour [13, 17].”
Boosting predictions of treatment success (2010, methodology): “Logic regression (9) is a technique that builds up Boolean rules based primarily on categorical variables and has seen considerable usage in genetic applications; other methods, such as the patient rule induction method (PRIM) (10) and extreme regression (11), are better suited for ordered variables”
Subgroup Identification and Variable Selection from Randomized Clinical Trial Data. (2013, methodology): “…one may instead wish to pre-define a statistical approach for identifying subgroups [Ruberg et al., 2010], which can be implemented using the data [Friedman and Fisher, 1999, Negassa et al., 2005, Su et al., 2008, 2009, Brinkley et al., 2010, Cai et al., 2011, Foster et al., 2011, Lipkovich et…”
Learning query-dependent prefilters for scalable image retrieval (2009, methodology): “We experimented with several algorithms for phrase selection, and describe one—a variant of “bump hunting” [ 3 ]—which has proved effective in practice.”
DroidDetector: Android Malware Characterization and Detection Using Deep Learning (2016, methodology): “We used the Patient Rule Induction Method (PRIM) bump hunting algorithm [30] which can learn a set of association rules and maximize a target variable of interest, to obtain the top ten 2-itemset association rules that have the maximum conﬁdence values, as shown in Table 3.”
Identifying and evaluating robust adaptive policy responses to climate change for water management agencies in the American west (2010, methodology): “To characterize the conditions that lead to this high-cost outcomes, we apply a bump-hunting algorithm called the “patient rule induction method” or PRIM [39] to identify the combination of values for a small number of uncertain factors that best explains those stateswhere the strategy performs poor”
Adaptive Hyper-box Matching for Interpretable Individualized Treatment Effect Estimation (2020, methodology): “, classication (e.g. Xu and Papageorgiou, 2009) and prediction (e.g. Goh and Rudin, 2014) but notably, not for causal inference (Khuat et al., 2019). These methods (and others, such as bump hunting, Friedman and Fisher, 1999) aim to nd adaptive boxes around individual data points and use MIP to nd b”
Evaluation Measures for Multi-class Subgroup Discovery (2009, background): “PRIM finds subregions of the instance space within which the value of the (continuous) output variable is considerably larger or smaller than its average value over the entire space.”