This study addresses the challenge of managing aging print collections in academic libraries by developing a data-driven framework that quantifies discipline-specific aging patterns and predicts future trends to inform dynamic shelving strategies.
We extend Bernal's document aging model to library circulation data. Utilizing loan records (2001–2023) from 22 disciplines at a major university library, we fit a negative exponential decay model to quantify aging half-lives. A Gradient Boosting Regression Trees (GBRT) algorithm is then employed to forecast future aging parameters. These predictions inform a novel four-quadrant classification system and a weighted hierarchical prioritization model for shelving.
The negative exponential model demonstrated robust goodness-of-fit (mean R2 = 0.75) across disciplines, validating its applicability to loan data. Significant cross-disciplinary heterogeneity was observed (mean half-life: 4.9 ± 0.7 years). The GBRT model achieved good prediction accuracy for initial loan rates (overall MAPE = 11.9%). The resulting framework categorizes disciplines into three actionable preservation tiers (Tier-1: open shelves; Tier-2: thematic displays; Tier-3: dense storage).
The proposed framework offers library managers a replicable, data-driven tool to optimize physical space allocation, balance collection utility with preservation needs, and justify resource reconfiguration.
This research makes three key contributions: (1) it theoretically extends document obsolescence analysis from citations to circulation data; (2) it methodologically innovates by integrating bibliometric modeling with machine learning for prediction; (3) it provides practitioners with a novel, predictive decision-making framework for evidence-based space management.
