[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85737-en":3,"doc-seo-85737-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},85737,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","Reliability-Aware Ensemble Classification Under Class Imbalance","Cervical cytology classification models are evaluated on class-balanced benchmarks, yet real-world liquid-based cytology (LBC) collections are typically small and class-imbalanced. This study performs a class-imbalance-aware, calibration-aware ensemble classification on the Mendeley LBC dataset using its four-class Bethesda taxonomy (NILM, LSIL, HSIL, SCC). Weighted sampling trains three lightweight models, while temperature scaling calibrates probabilities on a held-out subset. Calibration reduces expected calibration error, Brier score, and negative log-likelihood, with discrimination largely unchanged.","Reliability-Aware Ensemble Classification Under Class Imbalance: A Calibration Study on Liquid-Based Cervical Cytology  \nNisreen Albzour, Sarah S. Lam  \nSchool of Systems Science and Industrial Engineering, Binghamton University, Binghamton, NY 13902, USA  \n* [Corresponding author: ](Corresponding author: nalbzour@binghamton.edu)[nalbzour@binghamton.edu](Corresponding author: nalbzour@binghamton.edu)  \nORCID: [https://orcid.org/0009-0000-9317-340X](https://orcid.org/0009-0000-9317-340X)  \nAbstract  \nCervical cytology classification models are typically evaluated on curated, class-balanced benchmarks, but real-world liquid-based cytology (LBC) collections are often small and class-imbalanced. This paper presents a class-imbalance-aware and calibration-aware ensemble classification study on the Mendeley LBC dataset, using its native four-class Bethesda taxonomy (NILM, LSIL, HSIL, SCC) rather than a collapsed binary formulation. Three lightweight architectures (Swin-Tiny, TinyViT-5M, DenseNet121) are trained directly on Mendeley LBC using weighted random sampling to counteract class imbalance, and compared against two soft-voting ensembles (Hybrid-2, Hybrid-3). Post-hoc temperature scaling is fit on a held-out calibration subset carved out of the training portion of each cross-validation fold, distinct from both the training data used to fit model weights and the evaluation fold used for final metrics, avoiding the optimistic calibration estimates that result when the same data is used for both purposes. Calibration substantially reduces expected calibration error, Brier score, and negative log-likelihood for every model and ensemble configuration tested, while discrimination metrics (accuracy, macro-F1, macro-AUROC) remain essentially unchanged. Ensemble size shows no consistent additional reliability benefit over the best individual model once all configurations are properly calibrated. Confusion matrices show that all classification errors, across every configuration, are confined to the boundary between high-grade lesions (HSIL) and carcinoma (SCC); no errors involve the negative (NILM) or low-grade (LSIL) categories. These results suggest that, for this dataset, calibration is the dominant lever for reliability, not ensemble size, though this conclusion should be read in light of the dataset's modest size.  \n1. Introduction  \nAutomated classification of cervical cytology images has been studied extensively on curated benchmarks such as SIPaKMeD, where class representation is close to balanced by design. Real-world liquid-based cytology (LBC) collections rarely share this property: class frequencies reflect actual screening populations, in which high-grade and malignant findings are, by definition, rarer than negative findings. In one reported LBC screening program, for example, the negative (NILM) category accounted for roughly 89% of 1,293 samples while high-grade lesions accounted for well under one percent [1] . Models evaluated only on balanced benchmarks provide limited evidence about how they will behave when trained and deployed on this kind of skewed, real-world data.  \nThis paper studies the Mendeley LBC dataset directly, rather than as an external validation target for models trained elsewhere. Two aspects of this dataset motivate the specific focus of this study. First, its class distribution is naturally imbalanced, dominated by the negative (NILM) category, which requires explicit imbalance handling during training rather than accuracy-only evaluation. Second, prior work evaluating models on Mendeley LBC as an external-validation set has typically collapsed its native four-class Bethesda taxonomy into a binary Normal-versus-Abnormal task to reconcile it with a differently-labeled training source. Binary formulations are common in LBC deep learning more generally; whole-slide LBC screening models, for instance, are frequently trained to separate neoplastic from non-neoplastic specimens rather  \nthan to grade lesi","cbCaifCd0m1yLlEW","https://ap.wps.com/l/cbCaifCd0m1yLlEW","pdf",725537,1,11,"English","en",105,"# Abstract\n# Introduction\n# Related Work\n## Dataset and Cervical Cytology Classification\n## Ensemble Methods and Calibration (Overview)\n## Class-Imbalance Handling","[{\"question\":\"Why does the study focus on class imbalance in liquid-based cervical cytology?\",\"answer\":\"Real-world LBC datasets reflect screening populations where negative cases dominate and high-grade or malignant findings are much rarer. This makes class-balanced benchmark evaluation insufficient for understanding deployed behavior.\"},{\"question\":\"What calibration method is used, and how is it evaluated?\",\"answer\":\"Post-hoc temperature scaling is fitted on a calibration subset held out from both model-weight training and the final evaluation fold in each cross-validation iteration. This design avoids overly optimistic calibration estimates.\"},{\"question\":\"What is the main finding about reliability versus discrimination?\",\"answer\":\"Calibration substantially improves reliability metrics such as expected calibration error, Brier score, and negative log-likelihood across tested models and ensembles. Discrimination metrics (accuracy, macro-F1, macro-AUROC) remain essentially unchanged, and ensemble size provides no consistent extra reliability gain once properly calibrated.\"}]",1784205926,28,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"reliability-aware-ensemble-classification-under-class-imbalance","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/reliability-aware-ensemble-classification-under-class-imbalance/85737/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does the study focus on class imbalance in liquid-based cervical cytology?","Question",{"text":75,"@type":76},"Real-world LBC datasets reflect screening populations where negative cases dominate and high-grade or malignant findings are much rarer. This makes class-balanced benchmark evaluation insufficient for understanding deployed behavior.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What calibration method is used, and how is it evaluated?",{"text":80,"@type":76},"Post-hoc temperature scaling is fitted on a calibration subset held out from both model-weight training and the final evaluation fold in each cross-validation iteration. This design avoids overly optimistic calibration estimates.",{"name":82,"@type":73,"acceptedAnswer":83},"What is the main finding about reliability versus discrimination?",{"text":84,"@type":76},"Calibration substantially improves reliability metrics such as expected calibration error, Brier score, and negative log-likelihood across tested models and ensembles. Discrimination metrics (accuracy, macro-F1, macro-AUROC) remain essentially unchanged, and ensemble size provides no consistent extra reliability gain once properly calibrated.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]