Temporal Feature‐Integrated Algorithm for Anomaly Detection and Trend Recognition in Industrial Internet of Things Systems

ABSTRACT Accurate and timely monitoring of dynamic processes is essential for effective management and decision‐making in complex industrial internet of things (IIoT) environments. However, real‐world IIoT systems often experience unexpected deviations caused by factors such as resource misallocation, environmental fluctuations, and unforeseen operational disturbances. To address these challenges, this paper proposes TEMPRO (temporal progress recognition and outlier detection), an innovative algorithm that integrates temporal features for real‐time anomaly detection and trend recognition. TEMPRO is built upon an LSTM‐based temporal modeling backbone with an adaptive error‐triggered adjustment mechanism, enabling robust learning of long‐term dependencies under non‐stationary conditions. TEMPRO employs time‐series analysis combined with advanced machine learning techniques to detect deviations from normal operational patterns, capture evolving trends, and provide early warnings of potential disruptions. By leveraging historical temporal data from interconnected IIoT devices, TEMPRO can recognize both short‐term anomalies and long‐term trend shifts, offering actionable insights to support proactive and intelligent decision‐making. The performance of TEMPRO is validated through IIoT‐based case studies, demonstrating its effectiveness in enhancing system monitoring, improving operational stability, and enabling timely interventions in complex industrial environments.