Event-preserving feature engineering for intermittent demand forecasting using SHOS

B Sendhil Nathan, Veera Siva Reddy B, C Chandrasekhara Sastry, Sachin Salunkhe, Robert Cep · 2026

Read Event-preserving feature engineering for intermittent demand forecasting using SHOS by B Sendhil Nathan, Veera Siva Reddy B, C Chandrasekhara Sastry, Sachin Salunkhe, Robert Cep (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13521953.txt.

Description

B Sendhil Nathan and colleagues propose in 2026 a feature-building method called SHOS for forecasting spare-parts demand that arrives in rare, irregular bursts. Standard models drown in zeros, while fancy architectures often ignore careful input design.

Demand is split into chance of an order occurring and expected size when one occurs, smoothed with sparsity-aware weighting to preserve real events. Testing used about 56,000 dealer-part series from automotive aftermarket data with rolling checks. Tree models fed with the new features beat raw-feature versions on accuracy and stability across varied demand segments. Signal preservation analyses support the design choice.

Catalog record

CREATORB Sendhil Nathan, Veera Siva Reddy B, C Chandrasekhara Sastry, Sachin Salunkhe, Robert Cep
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE110054
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13521953

Browse the full library · About the Tower of Babel