Issues using logistic regression with class imbalance, with a case study from credit risk modelling

Li, Yazhe; Bellotti, Tony; Adams, Niall M.

doi:10.3934/fods.2019016

Cited by 11 publications

(7 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our goal is to predict this outcome based on loan and borrower features available at origination. This setup is consistent with Li et al [14], although our sample is much larger.…”

Section: High-sensitivity and High-specificity Regionssupporting

confidence: 93%

“…Theorem 4.4 (Exactly Exponential w). Suppose w has the form in (14). Suppose Conditions 3-5 hold, with k = 0 in (11).…”

Section: Convergence Under Infinite Imbalancementioning

confidence: 99%

“…Li, Belloti, and Adams [14] consider regularized logistic regression with an L 1 or L 2 penalty on β. They show that the optimal β N converges to zero under infinite imbalance.…”

Section: Convergence Under Infinite Imbalancementioning

confidence: 99%

See 2 more Smart Citations

Linear Classifiers Under Infinite Imbalance

Glasserman¹,

Li²

2021

Preprint

View full text Add to dashboard Cite

We study the behavior of linear discriminant functions for binary classification in the infinite-imbalance limit, where the sample size of one class grows without bound while the sample size of the other remains fixed. The coefficients of the classifier minimize an expected loss specified through a weight function. We show that for a broad class of weight functions, the intercept diverges but the rest of the coefficient vector has a finite limit under infinite imbalance, extending prior work on logistic regression. The limit depends on the left tail of the weight function, for which we distinguish three cases: bounded, asymptotically polynomial, and asymptotically exponential. The limiting coefficient vectors reflect robustness or conservatism properties in the sense that they optimize against certain worst-case alternatives. In the bounded and polynomial cases, the limit is equivalent to an implicit choice of upsampling distribution for the minority class. We apply these ideas in a credit risk setting, with particular emphasis on performance in the high-sensitivity and high-specificity regions.

show abstract

“…Our goal is to predict this outcome based on loan and borrower features available at origination. This setup is consistent with Li et al [14], although our sample is much larger.…”

Section: High-sensitivity and High-specificity Regionssupporting

confidence: 93%

“…Theorem 4.4 (Exactly Exponential w). Suppose w has the form in (14). Suppose Conditions 3-5 hold, with k = 0 in (11).…”

Section: Convergence Under Infinite Imbalancementioning

confidence: 99%

See 1 more Smart Citation

Linear Classifiers Under Infinite Imbalance

Glasserman¹,

Li²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…It is noteworthy that most of the reported studies are based on a combination of feature selection and relatively simple models such as Multivariate Logistic Regression (MLR) or Diagonalized Linear Discriminant Analysis (DLDA). However, because there is no clear correlation between the expression of a gene and a NAC response vector (mean PCC

), linear models may not be the appropriate choice for classifying the NAC response [ 28 ]. In addition, some studies have utilized prior marker information to restrict the search space of optimal markersets, but these studies only utilized the associated genes of predefined mechanisms such as the immune response [ 29 ] and the E2F pathway [ 19 , 20 ], which may not fully represent the mechanisms of the NAC response.…”

Section: Introductionmentioning

confidence: 99%

Development of Gene Expression-Based Random Forest Model for Predicting Neoadjuvant Chemotherapy Response in Triple-Negative Breast Cancer

Park

2022

Cancers

View full text Add to dashboard Cite

Neoadjuvant chemotherapy (NAC) response is an important indicator of patient survival in triple negative breast cancer (TNBC), but predicting chemosensitivity remains a challenge in clinical practice. We developed an 86-gene-based random forest (RF) classifier capable of predicting neoadjuvant chemotherapy response (pathological Complete Response (pCR) or Residual Disease (RD)) in TNBC patients. The performance of pCR classification of the proposed model was evaluated by Receiver Operating Characteristic (ROC) curve and Precision Recall (PR) curve. The AUROC and AUPRC of the proposed model on the test set were 0.891 and 0.829, respectively. At a predefined specificity (>90%), the proposed model shows a superior sensitivity compared to the best performing reported NAC response prediction model (69.2% vs. 36.9%). Moreover, the predicted pCR status by the model well explains the distance recurrence free survival (DRFS) of TNBC patients. In addition, the pCR probabilities of the proposed model using the expression profiles of the CCLE TNBC cell lines show a high Spearman rank correlation with cyclophosphamide sensitivity in the TNBC cell lines (SRCC =0.697, p-value=0.031). Associations between the 86 genes and DNA repair/cell cycle mechanisms were provided through function enrichment analysis. Our study suggests that the random forest-based prediction model provides a reliable prediction of the clinical response to neoadjuvant chemotherapy and may explain chemosensitivity in TNBC.

show abstract

“…It is precisely this feature that makes the DC regression more efficient with imbalanced classifiers. A recent study on imbalanced classifiers isLi et al (2019) Albert and Anderson (1984),Allison (2008),. and Ghosh et al (2018) study convergence issues.…”

mentioning

confidence: 99%

Identifying structural VARs from sparse narrative instruments: Dynamic effects of US macroprudential policies

Budnik

Rünstler

2023

J of Applied Econometrics

View full text Add to dashboard Cite

Summary We study identification in Bayesian proxy VARs for instruments that consist of sparse qualitative observations indicating the signs of shocks in specific periods. We propose the Fisher discriminant regression and a non‐parametric sign concordance criterion as two alternative methods for achieving correct inference in this case. The former represents a minor deviation from a standard proxy VAR, whereas the non‐parametric approach builds on set identification. Our application to US macroprudential policies finds persistent declines in credit volumes and house prices together with moderate declines in GDP and inflation and a widening of corporate bond spreads after a tightening of capital requirements or mortgage underwriting standards.

show abstract

Issues using logistic regression with class imbalance, with a case study from credit risk modelling

Cited by 11 publications

References 22 publications

Linear Classifiers Under Infinite Imbalance

Linear Classifiers Under Infinite Imbalance

Development of Gene Expression-Based Random Forest Model for Predicting Neoadjuvant Chemotherapy Response in Triple-Negative Breast Cancer

Identifying structural VARs from sparse narrative instruments: Dynamic effects of US macroprudential policies

Contact Info

Product

Resources

About