Continuous Risk-Based Scheduling for Food Safety Inspection: Integrating Historical Decay with Event-Driven Consumer Review Signals

Akriti Bhat; Wayne Zhu; Ishan Gupta

cs.AIOct 10, 2026

Continuous Risk-Based Scheduling for Food Safety Inspection: Integrating Historical Decay with Event-Driven Consumer Review Signals

Authors: Akriti Bhat, Wayne Zhu, Ishan Gupta

Abstract

Batch risk-based inspection scheduling struggles to detect a critical failure class: food facilities with recent clean inspections that subsequently develop severe violations. Traditional systems recompute risk on fixed planning cycles using static historical data, leaving blind spots between inspection cycles. To address this, we introduce the Continuous Inspection Risk (CIR) system, an event-driven scheduling pipeline that re-ranks the inspection queue dynamically as new data arrives. CIR combines a decay-weighted historical score with a stream of real-time consumer reviews triaged by a language-model classifier. A key structural contribution is our temporal partition rule, which strictly excludes reviews predating the most recent inspection to prevent superseded signals from skewing current risk. Evaluating CIR on 9,505 routine inspections in Santa Clara County, California (2024-2026), we show that history-based scoring alone reduces mean time-to-detection of major violations by 69 days (20.9%) under a 30% inspection budget, but fails to anticipate severe vermin-related closures (0/8 detected). Incorporating post-inspection review signals drastically improves performance, predicting 7 of 8 vermin closures with an average lead time of 117 days-firing on the exact day of review publication in four instances. Introducing a review corroboration rule improves precision from 0.70 to 0.78 without loss of recall. Finally, we analyze critical implementation challenges, documenting a data-hygiene pitfall (stale review leakage across inspection boundaries) and an evaluation limitation involving unobserved facility self-remediation. This work provides the first case-level evaluation of event-driven inspection triggering using open data and provides a reproducible benchmark for review-augmented public health systems

Figures & tables

	Predicted SAFETY RISK	Predicted NOT RELEVANT
Actual SAFETY RISK	16 (TP)	1 (FN)
Actual NOT RELEVANT	2 (FP)	11 (TN)

Table 1. Confusion matrix for the triage classifier on the 30-review hand-labeled validation set.

Panel of Figure 1 — Figure 1: Major-violation catch rates under three inspection scheduling policies. On 7,953 held-out routine inspections from Santa Clara County, chronological scheduling (M0) achieves an 89.7% catch rate at a 100% inspection budget, the score-threshold heuristic (M1b; prior score < 90) achieves 65.0% at a 36% budget, and the Layer 1 historical decay model (Magent) achieves 49.0% at a 30% budget.

Panel of Figure 2 — Figure 2: Mean time to major-violation detection under three inspection scheduling policies. On the same 7,953 held-out routine inspections, mean time to detection is 329.7 days for chronological scheduling (M0), 312.3 days for the score-threshold heuristic (M1b), and 260.7 days for the Layer 1 historical decay model (Magent). Magent reduces mean time to detection by 69 days (20.9%) relative to M0, with inspection budgets of 30% and 100%, respectively.

Scheduler	Budget	Catch Rate	Mean Days to Detection
M 0 (chronological)	100%	89.7%	329.7
M 1b (score < 90)	36%	65.0%	312.3
M agent (Layer 1 only)	30%	49.0%	260.7

Table 1. Layer 1 backtest on 7,953 held-out inspections.

Restaurant	Last Score	Layer 1 Flags It?	Closure Cause
Red Robin	91	No	Vermin
Pacific Catch	82	No	Vermin
Benihana	88	No	Vermin
Round Table Pizza	88	No	Vermin
Wienerschnitzel	82	No	Vermin
Black Bear Diner	82	No 1	Vermin

Table 2. Layer 1 outcome per case. 1 Even the three facilities with visibly sliding scores never reach the scheduling threshold on decay alone within the available window: at these score levels the decay term tops out well below it. Total: 0 of 8.

Panel of Figure 3 — Figure 3: Recall on the eight-case set by evidence configuration. History alone: 0/8. History plus pre-window reviews, evaluated correctly: still 0/8 - stale reviews describe conditions an inspector has since checked. Only the live layer moves the number: 7/8.

Configuration	Recall (of 8)	Precision 2	Mean Lead
Layer 1 only	0/8 (0.00)	N/A	-
Layer 1+3, naive single-review trigger	7/8 (0.875)	0.70	117 days
Layer 1+3, corroboration + severity exception	7/8 (0.875)	0.78	117 days

Table 3. Case-set results. n = 8 closures (positives), 6–7 comparison facilities (negatives). 2 Precision here measures agreement with closure labels; Section 8 explains why that is not the same thing as correctness.

Panel of Figure 4 — Figure 4: Days between the system’s flag and the actual closure, per restaurant. Mean 117 days across the seven catches. The one miss is a data-coverage gap, detailed below.

Panel of Figure 5 — Figure 5: For four of seven catches, one qualifying review moves the facility up the queue the day it is posted.

Parameter	Value / Description
$\alpha$ (score weight)	1.0
$\lambda$ (decay rate)	0.01 per day
$\beta$ (live review weight)	10 (sensitivity sweep, Section 7.1)
Scheduling threshold	70th percentile of risk score across test-set facilities
Classifier confidence threshold	0.75
Single-review discount	0.3 $\times$ by default; 0.6 $\times$ if confidence $\geq$ 0.95 and severity $\geq$ 8; 1.0 $\times$ if the last score is < 90 or there are $\geq$ 2 qualifying Layer 3 reviews.

Table 10

Sub-category	Assigned Severity
Illness / food poisoning report	9
Pest sighting (roach, rodent, insect)	8–9
Raw or undercooked food, contamination	7
Unsanitary premises (dirty surfaces, sewage odor)	6–7
Hygiene lapse (bare-hand contact, dirty utensils)	6
Mold or spoiled food	6

Table 11

Industrial Inspection

Industrial inspection uses sensing, computer vision, and robotics to detect, measure, and explain defects and verify product or infrastructure quality.

See all 120 papers

Aug 3, 2026

Leveraging AI for fine-grained food safety risk forecasting in sparse data conditions

Ensuring food safety represents a critical public health challenge, particularly when inspection resources are limited and regional sampling data are sparse. This study proposes a Transformer-based framework capable of forecasting fine-grained, city-level food safety risks by unifying over 11 million inspection records with supplemental demographic, economic, and environmental indicators extracted from the Statistical Yearbook. A three-stage pretraining design leverages partial supervision from the Wilson interval (capturing both safety and risk rankings), together with semi-supervised label refinement, to effectively utilize historical records even when local sample sizes are insufficient. Experimental evaluations on data from 2022 show that the proposed approach outperforms baselines significantly. A subsequent field experiment in collaboration with the Zhejiang Provincial Administration for Market Regulation further demonstrates improved detection rates and more efficient allocation of inspection resources compared to a manually developed plan. Observations of regulatory decision-making reveal a threshold-based heuristic employed by inspectors, hinting that additional training or decision-support interfaces could further enhance the impact of AI-generated risk scores. Overall, these findings underscore that a rigorous integration of large-scale public inspection data, Wilson interval-based confidence modeling, and advanced deep learning can facilitate earlier and more granular identification of food safety threats. By reducing reliance on reactive measures alone, the proposed framework has the potential to advance proactive, data-driven oversight of the global food supply.

Dongqi Wang, Weiwei Chen, Han Zhou +1

School of Management, Zhejiang University · Zhejiang Key Laboratory of Decision Intelligence, Zhejiang University · Department of Supply Chain Management, Rutgers University

Spatiotemporal Forecasting

Ask PDF

Sep 17, 2025

Risk-Calibrated Bayesian Streaming Intrusion Detection with SRE-Aligned Decisions

[Corrected v2: an audit found that the score, threshold, and latency descriptions below are not what the shared codebase implements, and that the evaluation streams are assembled constructions. See the correction note on the title page and the corrected companion work, arXiv:2605.24696 (corrected v3), artifact doi:10.5281/zenodo.22673735.] We present a risk-calibrated approach to streaming intrusion detection that couples Bayesian Online Changepoint Detection (BOCPD) with decision thresholds aligned to Site Reliability Engineering (SRE) error budgets. BOCPD provides run-length posteriors that adapt to distribution shift and concept drift; we map these posteriors to alert decisions by optimizing expected operational cost under false-positive and false-negative budgets. We detail the hazard model, conjugate updates, and an O(1)-per-event implementation. A concrete SRE example shows how a 99.9% availability SLO (43.2 minutes per month error budget) yields a probability threshold near 0.91 when missed incidents are 10x more costly than false alarms. We evaluate on the full UNSW-NB15 and CIC-IDS2017 benchmarks with chronological splits, comparing against strong unsupervised baselines (ECOD, COPOD, and LOF). Metrics include PR-AUC, ROC-AUC, Brier score, calibration reliability diagrams, and detection latency measured in events. Results indicate improved precision-recall at mid to high recall and better probability calibration relative to baselines. We release implementation details, hyperparameters, and ablations for hazard sensitivity and computational footprint. Code and reproducibility materials will be made available upon publication; datasets and implementation are available from the corresponding author upon reasonable request.

Michel A. Youssef

Independent Researcher · Independent Researcher, Lebanon

Cost-Sensitive Learning Probability Calibration Online Anomaly Detection Network Intrusion Detection

Ask PDF

Oct 7, 2026

PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift

Intrusion detectors can confidently misclassify attacks that were not seen during training. Human review can correct these errors, but only a limited number of cases can be checked. Uncertainty-based review may overlook confident errors, while anomaly scores alone do not show whether changing the review plan will correct more errors. We introduce PairAudit to find overlooked errors and improve review under a fixed budget. Its graph tokens capture prediction patterns across connected nodes. Rather than building another predictor through feature aggregation, PairAudit uses unusual relational patterns to uncover potential errors in existing predictions. Human feedback then helps decide whether these findings justify changing review priorities. Experiments across security tasks show that PairAudit corrects more errors on average than uncertainty-based review, including more errors on unseen attacks. These gains account for all review costs and do not require retraining the detector.

Jiran Tao, Binyan Jiang

The Hong Kong Polytechnic University

Graph-Based Anomaly Detection Network Intrusion Detection Black-Box Model Auditing Anomaly Detection OOD Detection

Ask PDF

Abstract

Industrial Inspection

Explore similar work

Leveraging AI for fine-grained food safety risk forecasting in sparse data conditions

Risk-Calibrated Bayesian Streaming Intrusion Detection with SRE-Aligned Decisions

PairAudit: Guiding Human Review with Graph Tokens under Distribution Shift