Hellinger distance decision trees are robust and skew-insensitive

By David A. Cieslak, T. R. Hoens, Nitesh V. Chawla, W. Philip Kegelmeyer Jr.. 2011

251 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

How researchers see this paper

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

In Artificial Intelligence

In Information Systems

Citations in the corpus

Box drawings for learning with imbalanced data (2014, methodology): “The works most similar to ours are that of the Patient Rule Induction Method (PRIM) (Friedman and Fisher 1999) and decision tree methods for imbalanced classiﬁcation (e.g., (Japkowicz 2003)), as they partition the input space like our work.”
Sequence classification for credit-card fraud detection (2018, methodology): “[Cieslak et al., 2012] proposed to use the Hellinger distance as splitting criterion in decision trees.”
Classification with class imbalance problem: A review (2015, methodology): “To address these issues, [110] discussed a novel decision tree algorithm named Hellinger Distance Decision Trees (HDDT) that apply Hellineger distance for splitting criterion.”
DeepSMOTE: Fusing Deep Learning and SMOTE for Imbalanced Data (2021, methodology): “issues with the algorithmic approach are developing novel skew-insensitive split criteria for decision trees [22], [23], [24], using instance weighting for Support Vector Machines [25], [26], [27], or modifying the way different layers are trained in deep learning [28], [29], [30].”
Adaptive Machine Learning for Credit Card Fraud Detection (2015, methodology): “1 Hellinger Distance Decision Trees Originally introduced to quantify the similarity between two probability distributions [233], the Hellinger Distance (HD) has been recently proposed as a splitting criterion in decision trees to improve the accuracy in unbalanced problems [103, 230].”
SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary (2018, background): “In the 1990s as more data and applications of machine learning and data mining started to become prevalent, an important challenge emerged: how to achieve desired classification accuracy when dealing with data that had significantly skewed class distributions (Sun et al., 2009; He & Garcia, 2009; Ló”
A Survey of Predictive Modeling on Imbalanced Domains (2016, methodology): “More recently, Cieslak et al. [2012] recommended the use of bagged HDDTs as the preferred method for dealing with imbalanced domains when using decision trees.”
Contrast Pattern-Based Classification for Bot Detection on Twitter (2019, methodology): “In these tables, we also show each tested combination based on patterns such as a method for mining contrast patterns (LCMine [49] or RFm [48]), a measure for evaluating a split criteria on a decision tree (Hellinger distance [57], Bhattacharyya [58], and information gain [59]), if it was filtered (”
Detecting Pneumatic Failures on Temporary Immersion Bioreactors (2016, methodology): “HeDex is a decision trees ensemble [19] using the Hellinger Distance [6] as a decision tree splitting criterion.”
A Pattern-Based Approach for Detecting Pneumatic Failures on Temporary Immersion Bioreactors (2019, methodology): “HeDex is based on a decision trees ensemble using the Hellinger distance [31] as a decision tree splitting criterion.”
Oversampling for Imbalanced Learning Based on K-Means and SMOTE (2017, methodology): “Combined with the Friedman test [17] and Holm’s method [24] , this evaluation method is also used by other authors working on the topic of imbalanced classification [8,19,41] .”
Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches (2013, background): “Therefore, they can be deﬁned as internal approaches that create new algorithms or modify existing ones to take the class imbalance problem into consideration [66,4,12,31,9].”
Learning from imbalanced data: open challenges and future directions (2016, unknown): “Recently a Hellinger-distance modification of decision trees [10] and neural networks [66] were proposed and proved to work highly efficiently.”