[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81771-en":3,"doc-seo-81771-105":30,"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},81771,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","Interpretable vs Learned Encoders for High-Cardinality Fraud Detection","Seven categorical encoders are benchmarked for fraud detection on the IEEE-CIS dataset (590,540 records, 3.5% positives, eight high-cardinality columns), using stratified 5-fold cross-validation with three repetitions. Five encoders are paired with frozen LightGBM to isolate encoder effects; CatBoost and TabNet provide cross-paradigm comparisons. Entity embeddings achieve the best AUC-ROC (0.9612), statistically tied with CatBoost (0.9602), and outperform tier group encoding. Target encoding is only 0.0023 worse while preserving auditor-friendly tier boundaries; TabNet fails under data scarcity. Per-column analysis attributes gains to joint multi-column representation.","Interpretable versus Learned Encoders for High-Cardinality Fraud Detection  \nXiao Han  \nGoizueta Business School Emory University Atlanta, GA, USA [xhan@alumni.emory.edu](xhan@alumni.emory.edu)  \nJingjing (May) Liu  \nElectrical Engineering and Computer Sciences University of California, Berkeley Kirkland, WA [mayliujj@berkeley.edu](mayliujj@berkeley.edu)  \nMoxuan Zheng  \nStern School of Business New York University New York, NY, USA [mz2156@nyu.edu](mz2156@nyu.edu)  \narXiv :2607 .00477v 1 [ cs .LG] 1 Jul 2026  \nZhen Zhang  \nSchool of Data Science University of Pennsylvania Philadelphia, PA, USA [billzhangzhen98@gmail.com](billzhangzhen98@gmail.com)  \nChenyu Wu  \nPratt School of Engineering Duke University  \nDurham, NC, USA  \n[wuchenyu999@outlook.com](wuchenyu999@outlook.com)  \nAbstract—A total of seven categorical encoding methods were tested on the IEEE-CIS fraud benchmark dataset (590,540 records, 3.5% positives, 8 high-cardinality columns). The encoders were evaluated using a stratified 5-fold cross-validation (CV) with three repetitions. Five of the encoders had identical frozen LightGBM learners in the downstream phase, allowing for controlled comparisons of their performance to eachother. CatBoost and TabNet were included as comparisons across paradigms using different learners. The entity embeddings produced the highest AUC-ROC (0.9612), with a statistically significant tie with that of CatBoost (0.9602) and statistically superior to tier group encoding (0.9548), whereas target encoding was only 0.0023 worse than tier group encoding and the auditorfriendly tier boundaries were maintained. Off-the-shelf TabNet did not outperform tree-based pipelines and collapsed under data scarcity. On AUC-PR, CatBoost leads (0.822 vs. 0.793); no encoder dominated both metrics. Per-column analysis confirmed the embedding advantage arises from joint multi-column representation.  \nIndex Terms—categorical encoding, fraud detection, entity embeddings, interpretability, tabular deep learning  \nI. INTRODUCTION  \nTabular transaction records contain many high-cardinality categorical variables (card identifiers, billing addresses, email domains, device fingerprints) with cardinalities from four to tens of thousands. How to represent such categories is an important modeling decision, but evidence is inconsistent: encoder benchmarks stay within the classical family [1], [2], while tabular deep-learning studies fix the encoding and vary the model [3]–[5] . A banking practitioner cannot determine whether neural entity embeddings [6] or a leakage-safe target encoder [7], [8] will serve best in a fraud model when modeltransparency requirements (such as SR 11-7 in the US, or analogous audit standards elsewhere) demand that scores be explainable.  \nThe question spans three axes. Accuracy margins are narrow, so isolating the encoder from the learner matters. Compute varies widely: end-to-end deep models can be an order of  \nmagnitude slower than tree-plus-embedding pipelines. Interpretability is critical under SR 11-7, which requires model-risk owners to justify scores, so a slightly less accurate but more interpretable encoder may be preferable. We report all three together.  \nWe address: what are the accuracy, interpretability, and computational trade-offs of seven encoders on a large fraud dataset with a fixed downstream learner, and where does lightweight statistical grouping sit on this continuum? Our contributions:  \n1) Cross-paradigm controlled experiment. We compared five encoders under a single fixed LightGBM learner on the IEEE-CIS dataset with stratified CV, plus CatBoost and TabNet as cross-paradigm reference points. Previous studies [3], [4] varied the model family with a default encoder; we varied only the encoder, finding ≈0.01 AUC-ROC attributable to encoding alone.  \n2) Auditor-readable encoding operationalized and benchmarked. We operationalized and benchmarked target-aware tier grouping (Bayesian-smoothed fraud rates binned into K ordi","cbCaijAJ05rNMOWJ","https://ap.wps.com/l/cbCaijAJ05rNMOWJ","pdf",267861,3,1,5,"English","en",105,"# Abstract\n# Introduction\n# Related Work\n# Methods","[{\"question\":\"Which encoding methods were evaluated and how was the comparison controlled?\",\"answer\":\"Seven categorical encoding methods were tested on IEEE-CIS using stratified 5-fold cross-validation with three repetitions. Five methods used frozen LightGBM learners so encoder differences could be isolated, while CatBoost and TabNet served as cross-paradigm references.\"},{\"question\":\"How do entity embeddings and tier grouping compare on AUC-ROC?\",\"answer\":\"Entity embeddings achieve the highest AUC-ROC (0.9612). They are statistically tied with CatBoost (0.9602) and outperform tier group encoding (0.9548).\"},{\"question\":\"What do the results imply for interpretability and regulated-audit requirements?\",\"answer\":\"The study frames interpretability, compute cost, and metric robustness as key outcomes under model-risk management expectations (e.g., SR 11-7). Target encoding preserves auditor-friendly tier boundaries, trading only a small AUC-ROC gap versus tier group encoding, while TabNet underperforms under data scarcity.\"}]",1784176055,13,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"interpretable-vs-learned-encoders-for-high-cardinality-fraud-detection","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/interpretable-vs-learned-encoders-for-high-cardinality-fraud-detection/81771/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-24","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},"Which encoding methods were evaluated and how was the comparison controlled?","Question",{"text":75,"@type":76},"Seven categorical encoding methods were tested on IEEE-CIS using stratified 5-fold cross-validation with three repetitions. Five methods used frozen LightGBM learners so encoder differences could be isolated, while CatBoost and TabNet served as cross-paradigm references.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How do entity embeddings and tier grouping compare on AUC-ROC?",{"text":80,"@type":76},"Entity embeddings achieve the highest AUC-ROC (0.9612). They are statistically tied with CatBoost (0.9602) and outperform tier group encoding (0.9548).",{"name":82,"@type":73,"acceptedAnswer":83},"What do the results imply for interpretability and regulated-audit requirements?",{"text":84,"@type":76},"The study frames interpretability, compute cost, and metric robustness as key outcomes under model-risk management expectations (e.g., SR 11-7). Target encoding preserves auditor-friendly tier boundaries, trading only a small AUC-ROC gap versus tier group encoding, while TabNet underperforms under data scarcity.","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":25},{"code":4,"msg":5,"data":92},[93,97,101,105,109,114,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":22,"slug":137},19,"General","general"]