Cost-effective prediction in medicine and marketing: only the difference between Bayesian model averaging and the single best-fit model

Darwen, Paul J. (2019) Cost-effective prediction in medicine and marketing: only the difference between Bayesian model averaging and the single best-fit model. In: Proceedings of the IEEE 31st International Conference on Tools for Artificial Intelligence. pp. 1274-1279. From: ICTAI 2019: IEEE 31st International Conference on Tools for Artificial Intelligence, 4-6 November 2019, Portland, OR, USA.

[img] PDF (Published Version) - Published Version
Restricted to Repository staff only

View at Publisher Website: https://doi.org/ 10.1109/ICTAI.2019.0017...
 
2


Abstract

There is a need for cost-effective tools to identify a small fraction of the population with a high probability of having a disease (in medicine) or buying the product (in marketing). Instead of the most probable model, Bayesian model averaging attempts to find the most probable prediction, given the data. None of these methods give satisfactory results in predicting from small-data, low-signal data sets. Using an example problem of predicting who has a rare lung cancer, this paper demonstrates that improved results can be found from considering only the difference between the predictions of the single best-fit model and Bayesian model averaging, and shows the correlation between precision for cancer and the size of that difference in predictions.

Item ID: 62315
Item Type: Conference Item (Research - E1)
ISBN: 978-1-7281-3798-8
Keywords: Bayesian model averaging, ensemble methods, committee of committees, cancer, medical diagnostics
Copyright Information: Copyright IEEE Computer Society
Date Deposited: 19 Feb 2020 03:34
FoR Codes: 42 HEALTH SCIENCES > 4202 Epidemiology > 420299 Epidemiology not elsewhere classified @ 10%
46 INFORMATION AND COMPUTING SCIENCES > 4606 Distributed computing and systems software > 460605 Distributed systems and algorithms @ 10%
46 INFORMATION AND COMPUTING SCIENCES > 4611 Machine learning > 461199 Machine learning not elsewhere classified @ 80%
SEO Codes: 92 HEALTH > 9202 Health and Support Services > 920203 Diagnostic Methods @ 70%
89 INFORMATION AND COMMUNICATION SERVICES > 8999 Other Information and Communication Services > 899999 Information and Communication Services not elsewhere classified @ 30%
Downloads: Total: 2
More Statistics

Actions (Repository Staff Only)

Item Control Page Item Control Page