EventGPT: Next-Event Pretraining on Banking Customer Histories
Abstract
A model pretrained to predict a customer's next event can estimate the probability of future events without training on their labels and provide a reusable representation for supervised downstream banking tasks. We pretrain EventGPT, a hierarchical causal transformer, by next-token maximum likelihood estimation on the public Multimodal Banking Dataset (MBD-mini: 100,224 clients, 38.8 million transactions and 66.3 million geolocation pings). Keeping one state per event, EventGPT trains in about two hours on one L4 GPU at windows up to 8,192 tokens. From its next-event distribution, with no labels, EventGPT scores five targets of the form "the client makes a transaction of a given kind for the first time next month" at ROC-AUC 0.79 to 0.92. Gradient boosting on 114 engineered features needs 10,000 to 300,000 labeled client-months to match it on four targets and stays below it on the fifth. For product prediction, frozen EventGPT embeddings score below the engineered features (mean ROC-AUC 0.787 against 0.808 over five test folds), and a rank-averaged ensemble of EventGPT, profile fields and token counts with the engineered features adds 0.014 to 0.017, of which 0.007 to 0.016 is EventGPT's own. The code and pretrained weights are released.
Related
- Mobile banking service quality and customer satisfaction in the Afghanistan banking sector through a modified E-SERVQUAL framework
- Why Customer Choice Models Matter
- The determinants of Open Banking adoption in the Moroccanbanking sector: an empirical validation of the UTAUT-OBMmodel
- Text Data Analysis and Classification Methods - Insights from Customer Letters in Life Insurance
- Measuring audience experience across 20 showings of the same theatre performance using customer feedback kiosks and a questionnaire
BibTeX
@misc{burakov2026eventgpt,
title = {{EventGPT: Next-Event Pretraining on Banking Customer Histories}},
author = {Denis Burakov},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.03302},
eprint = {2611.03302},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.03302},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - EventGPT: Next-Event Pretraining on Banking Customer Histories
AU - Denis Burakov
PY - 2026
DA - 2026/11/02
PB - arXiv
DO - 10.48550/arxiv.2611.03302
UR - https://arxiv.org/abs/2611.03302
AB - A model pretrained to predict a customer's next event can estimate the probability of future events without training on their labels and provide a reusable representation for supervised downstream banking tasks. We pretrain EventGPT, a hierarchical causal transformer, by next-token maximum likelihood estimation on the public Multimodal Banking Dataset (MBD-mini: 100,224 clients, 38.8 million transactions and 66.3 million geolocation pings). Keeping one state per event, EventGPT trains in about two hours on one L4 GPU at windows up to 8,192 tokens. From its next-event distribution, with no labels, EventGPT scores five targets of the form "the client makes a transaction of a given kind for the first time next month" at ROC-AUC 0.79 to 0.92. Gradient boosting on 114 engineered features needs 10,000 to 300,000 labeled client-months to match it on four targets and stays below it on the fifth. For product prediction, frozen EventGPT embeddings score below the engineered features (mean ROC-AUC 0.787 against 0.808 over five test folds), and a rank-averaged ensemble of EventGPT, profile fields and token counts with the engineered features adds 0.014 to 0.017, of which 0.007 to 0.016 is EventGPT's own. The code and pretrained weights are released.
ER -