[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-123616-en":3,"doc-seo-123616-105":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":4,"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":27,"seo_description":14,"update_tm":28,"read_time":29},123616,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",8,"Research & Report","Stabilizing machine learning models with Age-Period-Cohort inputs for scoring and stress testing","Machine learning credit scoring models often lose accuracy out-of-sample and out-of-time, and typical architectures inadequately incorporate economic scenarios needed for stress testing, cash-flow, or yield estimation. This research shows that adding Age-Period-Cohort lifecycle and environment functions improves out-of-sample robustness and enables both scoring and stress testing. The method is demonstrated for behavior scoring using account delinquency inputs, and evaluated in origination and behavior settings. Comparisons include multihorizon survival models and panel designs on prime US mortgage data.","TYPE Original Research PUBLISHED 08 June 2023  \nDOI 10. 3389/fams.2023.1195810  \nOPEN ACCESS  \nEDITED BY  \nRa􀀀aella Calabrese,  \nUniversity of Edinburgh, United Kingdom  \nREVIEWED BY  \nStefan Lessmann,  \nHumboldt University of Berlin, Germany Wouter Verbeke,  \nKU Leuven, Belgium  \n*CORRESPONDENCE  \nJoseph L. Breeden  \n [breeden@deepfutureanalytics.com](breeden@deepfutureanalytics.com)  \nRECEIVED 29 March 2023  \nACCEPTED 23 May 2023  \nPUBLISHED 08 June 2023  \nCITATION  \nBreeden JL and Leonova Y (2023) Stabilizing machine learning models with  \nAge-Period-Cohort inputs for scoring and stress testing.  \nFront. Appl. Math. Stat. 9:1195810 .  \ndoi: 10.3389/fams.2023.1195810  \nCOPYRIGHT  \n© 2023 Breeden and Leonova. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.  \nStabilizing machine learning models with Age-Period-Cohort inputs for scoring and stress testing  \nJoseph L. Breeden* and Yevgeniya Leonova Deep Future Analytics LLC, Santa Fe, NM, United States  \nMachine learning models have been used extensively for credit scoring, but the architectures employed su􀀀er from a significant loss in accuracy out-of-sample and out-of-time. Further, the most common architectures do not e􀀀ectively integrate economic scenarios to enable stress testing, cash flow, or yield estimation. The present research demonstrates that providing lifecycle and environment functions from Age-Period-Cohort analysis can significantly improve out-of-sample and out-of-time performance as well as enabling the model’s use in both scoring and stress testing applications. This method is demonstrated for behavior scoring where account delinquency is one of the provided inputs, because behavior scoring has historically presented the most di􀀈culties for combining credit scoring and stress testing. Our method works well in both origination and behavior scoring. The results are also compared to multihorizon survival models, which share the same architectural design with Age-Period-Cohort inputs and coe􀀈cients that vary with forecast horizon, but using a logistic regression estimation of the model. The analysis was performed on 30-year prime conforming US mortgage data. Nonlinear problems involving large amounts of alternate data are best at highlighting the advantages of machine learning. Data from Fannie Mae and Freddie Mac is not such a test case, but it serves the purpose of comparing these methods with and without Age-Period-Cohort inputs. In order to make a fair comparison, all models are given a panel structure where each account is observed monthly to determine default or non-default.  \nKEYWORDS  \ncredit scoring, survival models, Age-Period-Cohort, neural networks, stochastic gradient boosted trees  \n1. Introduction  \nMachine learning models are revolutionizing credit risk scoring. Models of common loan products for prime borrowers using credit bureau data have been re􀀂ned over decades, but novel products, alternate data sources [1–5], and lending to underserved populations have demonstrated the exceptional power of machine learning algorithms. However, the structural setup of these new machine learning models largely follows that of traditional cross-sectional logistic regression models – rank-ordering risk of default or a similar end state during a 􀀂xed outcome period. As lending becomes ever more competitive, this paradigm has several short-comings: rankings are not probabilities of default as needed for loan pricing, credit scores may erroneously explain trends in the in-sample data with score factor trends instead of macroeconomic trends, and credit risk rankings frequen","cbCaieQeKHOjrPs3","https://ap.wps.com/l/cbCaieQeKHOjrPs3","pdf",1451993,1,13,"English","en",105,"# Introduction\n## Motivation for stabilizing out-of-sample and out-of-time performance\n## Limits of cross-sectional logistic-style risk rankings\n## Age-Period-Cohort integration with machine learning and survival baselines","[{\"question\":\"Why do common machine learning credit scoring models degrade out-of-time?\",\"answer\":\"They often mirror cross-sectional logistic-style architectures that do not capture how macroeconomic conditions evolve, leading to ranking and probability calibration issues when environments shift.\"},{\"question\":\"What is the main idea behind using Age-Period-Cohort (APC) inputs?\",\"answer\":\"Lifecycle and environment functions derived from APC analysis provide structured effects so the scoring component aligns with known time-varying influences, improving stability out-of-sample and out-of-time.\"},{\"question\":\"How is the proposed approach validated and compared?\",\"answer\":\"The study trains models on prime 30-year US mortgage data with a monthly panel structure to observe default versus non-default, and compares against models without APC inputs and against multihorizon survival models.\"}]","Stabilizing machine learning models with Age-Period-Cohort inputs for scoring and stress testing | PDF",1785817652,33,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"stabilizing-machine-learning-models-with-age-period-cohort-inputs-for-scoring-and-stress-testing","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"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":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/stabilizing-machine-learning-models-with-age-period-cohort-inputs-for-scoring-and-stress-testing/123616/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-04",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why do common machine learning credit scoring models degrade out-of-time?","Question",{"text":75,"@type":76},"They often mirror cross-sectional logistic-style architectures that do not capture how macroeconomic conditions evolve, leading to ranking and probability calibration issues when environments shift.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the main idea behind using Age-Period-Cohort (APC) inputs?",{"text":80,"@type":76},"Lifecycle and environment functions derived from APC analysis provide structured effects so the scoring component aligns with known time-varying influences, improving stability out-of-sample and out-of-time.",{"name":82,"@type":73,"acceptedAnswer":83},"How is the proposed approach validated and compared?",{"text":84,"@type":76},"The study trains models on prime 30-year US mortgage data with a monthly panel structure to observe default versus non-default, and compares against models without APC inputs and against multihorizon survival models.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"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":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":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]