Answers are grounded in this paper and cite their sources.
Order-preserving (OP) indexes compact time-series windows that share relative-order structure, whereas forgetting-aware OP mining has mainly been studied as threshold-driven single-shot mining. We study a build-once/query-many problem: repeated global top-K ranking over a fixed weak-order index under endpoint exponential forgetting. A direct product-form factorization is algebraically correct but can create a binary64 intermediate in which an old-occurrence mass underflows while a positive depth exponential overflows. We derive a shift-safe factorization anchored at the rightmost occurrence start, so all evaluated exponential arguments are nonpositive and the rightmost-start term is exactly one. The resulting equivalence-stream query layer, EStream-OPF, formalizes canonical weak-order identity, maximal occurrence-equivalent streams, a deterministic machine comparator, rounded-score plateaus, K-head filtering, and comparator-consistent multiway merge. A Java 21 artifact implements the proposed method (B6) and five fair baselines/ablations. Across ten real benchmark instances from nine public data products, with series length n up to 500,000, B6 has lower median latency than the per-locus capacity-K heap baseline (B2) on 8/10 instances at maximum depth $L=12$, $K=50$, and half-life $n/4$; B2/B6 ratios include 1.72× for weekly CO₂, 1.69× for SILSO daily sunspots, 1.64× for UCI Household 100k, and 1.43× for NASDAQ. Factorization-only B4 is fastest on four datasets and stays within 6% of B6 on the four rows where B6 is fastest; Household 500k is at practical parity with B2. Reusing an index raises the build-amortized B2/B6 ratio from 1.01× for one query to 1.39× for 1,000 queries. The contribution is therefore conditional: benefits depend on occurrence-equivalent compression, the ratio $K/E$ of requested results to eligible streams, and index reuse. The K-head layer is optional once the stream mass is shared.
No takes yet. Share an insight, caveat, or question.
Nguyen et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: